You must’ve heard the “why is my washing machine sending soo much data to cloud?” joke from a few years ago.

Well, now replace ‘washing machine’ with ‘bed’

Quoting Michael Zimmermann

Is 16+gb/mo a normal amount of telemetry? Can you not do any local compute of “get hot” or “get cold” with a multi core processor and multiple gigabytes of memory? Can’t just repeat the previous nights settings?

It’s bad enough that you slapped a $ 200/yr subscription on things, worse that it doesn’t work at all without internet.

This whole thing came to light after many instances of EightSleep mattresses (pods) stopped functioning due to an AWS outage (which is hilarious and horrific at the same time).

This reminds me of a joke (and a truth):

If you want to know how dependent something is to the world, turn it off

A single high-value region on AWS went down recently and several external entities were disrupted and EightSleep pods were one of them.

EightSleep is fine but anything else that’s very critical is not.

This global incident should be a wake up call for dependent services to plan and support failovers. For consumer products to support offline mode. (either ways for a temp controlled bed to depend on cloud to function normally is ridiculous to a layman).

i’m referencing a post that can be found here .

I tried to make a joke on x dot com

~ updated at: 2025-10-22T04:28:05.994Z

feature, not a bug

All on-call engineers should be hooked up to @eightsleep pods whose functioning is dependent on current uptime of the services they are up-keeping :)

~ updated at: 2025-10-22T04:25:52.737Z

DHH on freedom and software:

Omarchy, Rails, Kamal, Hotwire, and all the rest of it doesn’t have to return anything on the investment beyond the joy of building something awesome and sharing it.

I have no interest in commercializing any of this.

i’m referencing a post that can be found here .

~ updated at: 2025-10-22T03:27:12.367Z

The browser wars have officially begun and whoever participates is forced to improve the agentic flows and there would be (IMO) a local maxima beyond which prompt injection becomes a catastrophic vector.

If google chooses to not participate, a significant portion of people can choose to migrate off google chrome and google might lose a significant chunk of market share or something idk… only until things go belly up when really bad prompt injection events come to light

If google chooses to participate, then it’s over. Google chrome has a large market share and this means google will also be cornered into the aforementioned local maxima until damage is already done to small 0f affected users. small ut large absolute number.

Prompt injections cannot be stopped until something novels comes about. but for time being it’s not looking good for google (in near term atleast). happy to be wrong about it.

i’m referencing a post that can be found here: https://x.com/daniellockyer/status/1980895459670634933?s=12

~ updated at: 2025-10-22T13:25:10.993Z

I found Guido Van Rossum casually answering python queries on X

~ updated at: 2025-10-20T02:06:15.205Z

I don’t particularly like resource contention irl

~ updated at: 2025-10-20T02:07:28.743Z

Today I learned that the generic function with multiple implementations based on the types of the a single arguments are called as single dispatch functions.

I’m dropping a reference to the mention on the official python 3 documentation

~ updated at: 2025-10-20T04:40:13.342Z

“… It’s to my detriment because sometimes my speaking thread out-executes my thinking thread”

I too speak faster than I think and this is similar to explanation I give to my peers 😂😭😭

https://x.com/karpathy/status/1979644538185752935?s=46

~ updated at: 2025-10-19T06:30:02.784Z

Shipping with assertion turned on, should ideally be done from the projects’ inception and not when project is already being used by customers.

The point I’m trying to make is that if you and you’re team are still working on fast iteration cycles and aren’t there yet, you can slowly adapt to using asserts in production code, and not all at once.

Turning asserts on in production codebase is definitely a way to help team build fast iteration cycles in terms of Nth order effects.

They learn (or) you go belly up. You choose. It’s a do or die thing.

~ updated at: 2025-10-17T08:50:20.963Z

Manifesting cost control with abundance mindset.

~ updated at: 2025-10-13T06:30:02.695Z

I don’t particularly like resource contention irl

~ updated at: 2025-10-11T09:40:37.804Z

Imo what makes Cursor great is using it as an additional tool vs a replacement - it’s rare you will one shot anything, that you can efficiently have an agent write all of the code for you

Things like typeahead or using the agent to do boilerplate generations is key

Typing manually is layer 1 and boilerplate generation is the prime subject to building additional layers.

Code generation tools are just a convenience layer to help when we’re constrained on resources like time, brain power etc.

Code generation can be used to automate testing boilerplate and test case transcription.

The test cases themselves could ideally be handwritten or designed to keep the context in the operators’ flow states.

https://x.com/zeeg/status/1976026560374485419?s=46

I really like terry tao’s take on red team vs blue team and how llms can be used effectively as red team agents.

~ updated at: 2025-10-09T12:30:05.408Z

TIL How to open any file in browser from the terminal

I always knew we could open any file in the browser on my Mac but I didn’t know how. Today I did and it’s super simple.

open -a Safari notes.txt

It’s the same for any file. Here’s an example of a markdown file:

open -a Safari notes.md

Side note

This knowledge opens a way to better deal with situations in real life. Like, if we have a markdown previewer extension on the browser of our choice, then it’s super useful during demos and presentations where we can keep the text editor and preview windows decoupled from each other.

Ideally, we don’t want to have two applications but have a single app showing editing and previewing mode.

Visual Studio Code is a great alternative

The existing solution is to use Visual Studio Code. with the inbuilt preview markdown feature. Command + Shift + V to preview the existing markdown file. It’s great for most use-cases.

If you need a simpler setup, one can use Neovim plugins. They are generally faster than electron app based solutions (not that it matters a lot but the inner purists in use can disagree).

~ updated at: 2025-10-09T07:54:47.495Z

Start by copying and you end up with some new entirely.

~ updated at: 2025-10-04T05:39:56.416Z

I fell into a rabbit hole on Twitter.

I was scrolling and I found a footage of how the maker of prince of persia rotoscoped VHS recording of his younger brother into the game mechanics.¹

I then found a link to a behind the Scenes sort of thing in retrospect²

“Your initial vision is just the first draft. You discover things along the way”

The designer Jordan kept a journal ( a log if you will) on every piece of inspiration or idea that he got from either media (or) just general feedback that his colleagues and friends gave while he was showing it to them.

“Whenever you are faced with two choices, through two voices, ask youself, which one speaks towards the overall picture” ~ paraphrased version

Reference:

~ updated at: 2025-10-04T05:19:46.410Z

One doesn’t simply get handed fast feedback loops off the bat.

They have to build one for themselves.

~ updated at: 2025-10-04T06:30:02.888Z

Excerpts from Signal’s post in response to E.U.’s Chat Control Policy Proposal¹

We are alarmed by reports that Germany is on the verge of a catastrophic about-face, reversing its longstanding and principled opposition to the EU’s Chat Control proposal.

Under the guise of protecting children, the latest Chat Control proposals would require mass scanning of every message, photo, and video on a person’s device, assessing these via a government-mandated database or AI model to determine whether they are permissible content or not.

These proposals ignore the strategic importance of private communications, and the longstanding technical consensus that you cannot create a backdoor that only lets the “good guys” in.

Signal shares a research paper on the dangers of client side scanning²

we were given a choice between building a surveillance machine into Signal or leaving the market, we would leave the market.

We urge Germany to be wise and to stand firm in its principles. We cannot let history repeat itself, this time with bigger databases and much much more sensitive data.

References:

~ updated at: 2025-10-04T02:40:15.003Z

The type of person to make the jump without the rope.

~ updated at: 2025-10-03T12:30:02.796Z

Veritasium posted another all timer on their channel called “Something Strange Happens When You Trace How Connected We Are”¹

This piece helps answer some questions I’ve been pondering about for a while now. Really well made video.

👏 Kudos to Derek and the entire team!

Now, I need to go deeper into Network Sociology and Emergent Behaviours to get my answers.

References:

~ updated at: 2025-10-01T08:30:05.093Z

How you write, should it affect how you think?

~ updated at: 2025-09-26T15:08:40.588Z

Getting fit isn’t rocket science but the execution requires rocket science level of determination and mindset:

1. 500-800 calories deficit
2. Minimum 150g protein
3. Lift full body 3x / week
4. 10K step a day, non-negotiable

That’s what we do inside our crew.

source on x dot com

biggest hurdle is a significant dependency on motivation and dependency is tech-debt.

~ updated at: 2025-09-25T05:50:45.881Z

Sometimes I say things so profound that people that it’s stupid. And sometimes I say things so stupid that people thinks it’s profound.

reference: https://twitter.com/twitter/status/1971120497267180008/

~ updated at: 2025-09-25T12:18:05.376Z

I think technology has transformed humans to edge nodes unintentionally (Nth Order Effect)

~ updated at: 2025-09-23T03:14:59.314Z

Listening to the ‘tenet’ album while driving to work and I’m not going to tell you if it’s a good or a bad idea (I think you already know)

~ updated at: 2025-09-17T05:17:11.747Z

I was today years old when I learned that Atom and R.S.S. feeds aren’t really the same thing.

Here’s what I learned

~ updated at: 2025-09-08T17:30:02.743Z

Think, Plan and then let A.I. execute.

Tell it how you think. Then it will teach you how to think better.

This helps tackle some “frustrations” I’ve encountered in the past.

~ updated at: 2025-09-05T17:30:05.445Z

I’ve been a fan of Founder’s podcast ever since my first listen.

I think it’s all in the delivery, where repetitions and redundancy helps a lot in keeping things in context.

The latest episode on Elon Musk was really inspiring. It must’ve been hard to make but the end result are worth listening to many many times.

Kudos to David Senra 👏🏻 for keeping the quality top-tier.

You can listen to ‘How Elon Works’ episode on Apple Podcasts

~ updated at: 2025-09-05T10:54:14.861Z

I have always used the voice dictation feature on the iPhone cuz I dislike typing on the phone, but I’ve never thought of enabling it on my MacBook.

Well one can easily guess how it feels now that I’ve enabled that feature.

What blew my mind is that it works even with neovim text editor (decently enough).

This is best for journaling and maintaining work logs.

~ updated at: 2025-09-05T10:01:57.485Z

Try to optimize for lowering context switching per scroll action.

this was intended to be a note to self, but also…

~ updated at: 2025-09-05T14:00:33.906Z

“….starting to use Postgres now after 12 years of scaling fine with SQLite!”

full thread for context from pieter levels on x dot com

~ updated at: 2025-09-04T02:11:32.429Z

Some interesting quotes:

My mind fills up with a few key projects and that’s it. I’m absorbed by those. That’s where my attention is.

If your mind isn’t filled with 2-3 key projects where you can riff your attention towards new things every day, then you’re doing something wrong.

Time and attention aren’t the same thing. They’re barely related.

They maybe two different dimensions but they’re definitely intertwined like Space-Time & Electro-Magnetic waves etc.

full reference on jason’s hey blog: https://world.hey.com/jason/the-difference-between-time-and-attention-bdd955eb

~ updated at: 2025-09-02T06:30:05.514Z

Today, I learned that you can format SQL statements in Postico 2 without worrying too much. There is an in-built feature in the query editor that allows you to do so.

Just select the SQL query, you want to format and right-click and choose Format SQL statement option.

Alternatively, we could also choose the statement and hit control F.

although, I’d prefer to not use such keyboard shortcuts because it could get dramatically problematic by bringing muscle memory to all this.

~ updated at: 2025-09-02T10:03:01.222Z

The sooner you test ideas, the more time you waste holding on to it before you realise it wasn’t the best of the ideas.

The best ideas come with a significant volume of lame or laughable ones.

~ updated at: 2025-09-02T06:30:07.998Z

we’re soo back!

“Huge day for 0.00050f followers” ~ Pewdiepie

pewd’s here might’ve thought “it’s a small thing, lemme just make a GitHub account and start pushing my files” but this could be a beginning of a beautiful (and sometimes unforgiving) rabbit-hole of things that bring joy to so many.

and yes, here’s a link Pewd’s GitHub account btw. Be kind :P

~ updated at: 2025-09-02T13:31:12.628Z

Many current programmers remember the Moore’s law curve. That was the period from the 70s through the 90s where clock rates and memory doubled every year or so. But that doubling stopped in 2005. Memory continued to get denser for a while, but even that has stopped. The current way that we increase computing power is by doubling the number of GPUs in server farms every year or so. But that can’t continue much longer. The power requirements, the space requirements, the capital requirements, and the sheer mass of it all, will exceed feasibility within a few more doublings. So don’t expect AIs to continue to improve exponentially. They’re headed for a brick wall, and that brick wall is not that far away.

This is definitely something to think about by zooming out. However, one must be optimistic towards (and prepare for) new scaling laws towards A.S.I. and beyond.

You can checkout the original post by Uncle Bob on x dot com .

~ updated at: 2025-09-02T13:38:35.697Z

I came across a low-key diabolical quote:

“Hard work may pay off in the end, but procrastination pays off now.” ~ Neil Fraser

~ updated at: 2025-09-01T11:51:42.427Z

It works well to not give 3rd party algorithms root access of your brain.

~ updated at: 2025-09-01T12:30:02.933Z

The sooner you test ideas, the more time you waste holding on to it before you realise it wasn’t the best of the ideas.

The best ideas come with a significant volume of lame or laughable ones.

~ updated at: 2025-09-01T10:19:07.282Z

What is the loss function you’re optimizing for in this day and age, fellow traveller?

~ updated at: 2025-08-31T06:30:02.890Z

Go for it. Don’t be afraid. Nobody cares. Even if they do, they’re just people.

source: https://x.com/melqtx/status/1961264778485387429

~ updated at: 2025-08-30T12:03:58.475Z

Perhaps the meaning of life is just to minimize context-switch per scroll function in the roaring twenties.

~ updated at: 2025-08-30T13:52:49.223Z

“I get it; ideas are hard to come by. If you let the X timeline brain rot you, you’ll have same 3 ideas everyone has else has.

I’ve been dealing with similar thoughts on social media in general.

Social media like X with algorithmic feed can end up becoming echo chambers. It’s not bad or wrong.

Suhail shares this sentiment after opening the following rant.

we do not need another loveable clone

Why is the world configured in such a manner?

These are not inherently startup ideas but if you can ask why and follow thread long enough you will find broken opportunities everywhere.

True.

Most people don’t even ask why.

Many do actually but the trail is left sooner than later onto the next thing.

you can check the original post yourself: https://x.com/Suhail/thread/1961636664599875913

~ updated at: 2025-08-30T11:05:45.854Z

you can like just just postpone worry indefinitely.

references https://x.com/bryan_johnson/status/1960425544023236701

~ updated at: 2025-08-29T12:30:05.436Z

Better to get your dopamine from improving your ideas than from having them validated.

source https://nat.org

~ updated at: 2025-08-29T12:36:31.972Z

Maybe just maybe the meaning of life is getting the next idea and learning (ideally the hard way) that it’s not that good. Because it’s stochastic knowledge that most ideas a not that good.

So sooner you learn why your next idea isn’t good enough the sooner you land on a really good one.

This thought was influenced by a note from John Carmack.

~ updated at: 2025-08-29T07:10:16.267Z

If you’re not able to communicate your state of mind about something in a RESTful way, you’re likely doing it wrong.

Because this could lead to jarring results rooting back to this miscommunication incident.

People arguing that sharing written pieces is the way to prevent this because being verbose and redundant in somethings works really well but the point is we gotta ensure the information reaches and is processed in their brains the way you intended it.

So written pieces make it easy for the author to place it all on the table but it doesn’t guarantee that the reading happens the way you intended it to.

This can be fixed with technology. Introducing multi-modality. A layer on top of the written piece.

A video of you explaining it the way you wanna intend it. Or even an audio piece.

If writing is layer 1, then video, audio can be a layer 2 and layer 3. Looking at problems like this can help appreciate technology more and more everyday.

NotebookLM product sort of attempts to solve this but nothing is as good as the original author attempting to convey the same thing through many modalities.

~ updated at: 2025-08-28T12:30:07.853Z

I am convinced that most inefficiencies are due to a lack of robust setup and teardown routines or their implementations.

~ updated at: 2025-08-28T15:30:02.810Z

so A.G.I. is just a database with agency

~ updated at: 2025-08-25T07:48:02.942Z

I’ve found that technology is best used to help develop (or even change) habits.

The process of unlearning can also be something technology can make accessible.

~ updated at: 2025-08-25T12:30:05.528Z

what is something no one has ever asked chatGPT, if asked it would unravel something drastically novel?

~ updated at: 2025-08-25T12:30:02.921Z

so A.G.I. is just a database with agency

~ updated at: 2025-08-24T08:46:30.387Z

Excerpt from contributing guidelines from Ghostty repository:

In a perfect world, AI assistance would produce equal or higher quality work than any human. That isn’t the world we live in today, and in most cases it’s generating slop. I say this despite being a fan of and using them successfully myself (with heavy supervision)!

via GitHub

~ updated at: 2025-08-20T06:30:05.330Z

Ah! NotionHQ finally launches offline mode.

https://x.com/notionhq/status/1957820181197000908?s=46

~ updated at: 2025-08-20T06:30:02.762Z

I’ve implemented individual post archival per day so each post can be linked to independently.

@GoHugoIO is amazing for post archival management.

https://tnvmadhav.me/feed/2025/08/17/

~ updated at: 2025-08-18T03:01:52.351Z

If you can’t recreate workflows from memory, you still haven’t done it enough.

It’s crazy that this has to be made explicit. This is a note to self btw.

🔗 You can read my other content pieces:

https://twitter.com/twitter/status/1956966725578080277

~ updated at: 2025-08-17T12:30:05.260Z

a man who is perpetually in the arena, will eventually act like there is nothing to lose.

~ updated at: 2025-08-17T15:05:33.259Z

On Using A.I. for actualy invention and discovery:

Most of the hype is about A.I. to eventually make more money, but the ultimate steelman has always been remove hurdles of group-think and instead generate unique ideas (with help of A.I.).

A perspective that isn’t yet getting enough attention: a way in which AI progress will soon deeply benefit the world is through the discovery and production of new technology.

We measure human progress by technological revolutions; hard to internalize what it’d mean to have a computer which can do much of the work for breakthroughs.

Quoting Greg Brockman via X

~ updated at: 2025-08-17T10:26:50.299Z

Bash scripts are great, if lesser than 30-50 LoC.

~ updated at: 2025-08-17T13:33:32.891Z

You can use github codespaces to bootstrap a piece of reproducible behaviour in a particular repository.

This is quite great for showcasing a piece of software. You can also run the software as if you’re running it locally without actually running it locally.

This includes GitHub copilot and other plugins.

I was more curious about GitHub Codespaces in general so I asked Claude Sonnet 3.5 that was within my Github Codespace how it all worked:

How It Works

When you create a codespace, GitHub: Spins up a container using your devcontainer.json configuration Sets up the development environment with specified tools and extensions Clones your repository into the container Connects it to VS Code (web or desktop)

“Code from anywhere with a browser”

Like for instance you can run, test and review code properly as you would from a pc/laptop.

“No local setup needed”

While this is technically true, but in practice, one would have to be well-versed with setup commands to make effective use of GitHub workspaces. Not an issue for the initiated.

Another major concern I had while learning about Codespaces was would the data be lost if code changes weren’t commited and pushed to remote source.

But looks like this concern was now relieved. The whole selling point of GitHub Codespaces was that a single codespace persists between sessions, meaning if you accidently close the tab “changes” can be restored.

However there are caveats!

If you’re personal account codespace wasn’t touched or used in last 30 days it will wipe the codespace clean.

While it’s a bummer, most of the workflows may not be hindered because almost everyone has a personal computer with them at most times. Codespace would mainly be used as a convenience layer to work out of a low computer hardware like a phone or tablets like Ipads.

Me personally, I like to make small commits and push to remote if it’s my personal projects. Also since this is remote, this is akin to having a VPS with OS and dependencies pre-installed for free to perform some cool experiments.

Moreover, it’s perfect for code-reviews.

All changes can be tested on a remote server without worrying about affecting your local server persistence or compute states.

Run a test, write a spec, edit typos, “quick fixes” and many more.

At least all of this can now be done without carrying your laptops to that park or cafe, if you have your tablets or iPads at your disposal.

Via Simon Willison’s TIL post on Configuring GitHub Codespaces

~ updated at: 2025-08-17T10:23:56.679Z

real shopping 🛒 but in V.R. 🥽

was this ever attempted? if not can it be attempted now? with A.I. helping with generating aisles in real-time.

like at @ikea

~ updated at: 2025-08-17T13:00:02.748Z

Killua going back and one-shotting the hunter exam during the greed island arc speaks volumes and is something I think about more often than not.

~ updated at: 2025-08-16T18:00:02.928Z

On the future of client-side A.I.

Traditional operating systems and client-side applications could be replaced so the compute can be saved for client side models instead.

The phone/computer will just become an edge node for AI, directly rendering pixels with no real operating system or apps in the traditional sense. There isn’t enough bandwidth to transmit video to all devices from the servers and there won’t always be good connectivity, so there still needs to be significant client-side AI compute.

Quoting Elon Musk via X

~ updated at: 2025-08-16T17:00:02.682Z

You can use github codespaces to bootstrap a piece of reproducible behaviour in a particular repository.

This is quite great for showcasing a piece of software. You can also run the software as if you’re running it locally without actually running it locally.

This includes GitHub copilot and other plugins.

I was more curious about GitHub Codespaces in general so I asked Claude Sonnet 3.5 that was within my Github Codespace how it all worked:

How It Works

When you create a codespace, GitHub: Spins up a container using your devcontainer.json configuration Sets up the development environment with specified tools and extensions Clones your repository into the container Connects it to VS Code (web or desktop)

“Code from anywhere with a browser”

Like for instance you can run, test and review code properly as you would from a pc/laptop.

“No local setup needed”

While this is technically true, but in practice, one would have to be well-versed with setup commands to make effective use of GitHub workspaces. Not an issue for the initiated.

Another major concern I had while learning about Codespaces was would the data be lost if code changes weren’t commited and pushed to remote source.

But looks like this concern was now relieved. The whole selling point of GitHub Codespaces was that a single codespace persists between sessions, meaning if you accidently close the tab “changes” can be restored.

However there are caveats!

If you’re personal account codespace wasn’t touched or used in last 30 days it will wipe the codespace clean.

While it’s a bummer, most of the workflows may not be hindered because almost everyone has a personal computer with them at most times. Codespace would mainly be used as a convenience layer to work out of a low computer hardware like a phone or tablets like Ipads.

Me personally, I like to make small commits and push to remote if it’s my personal projects. Also since this is remote, this is akin to having a VPS with OS and dependencies pre-installed for free to perform some cool experiments.

Moreover, it’s perfect for code-reviews.

All changes can be tested on a remote server without worrying about affecting your local server persistence or compute states.

Run a test, write a spec, edit typos, “quick fixes” and many more.

At least all of this can now be done without carrying your laptops to that park or cafe, if you have your tablets or iPads at your disposal.

Via Simon Willison’s TIL post on Configuring GitHub Codespaces

~ updated at: 2025-08-16T07:55:28.817Z

looking up your own guide from a few years ago to do something is very satisfying.

~ updated at: 2025-08-12T17:30:02.503Z

I’d like to believe that the intended purpose of multimodality is to invoke a cached catharsis if you will.

~ updated at: 2025-08-12T14:30:02.650Z

The internet runs because of a couple of goto statements in the lower levels of technological stack.

~ updated at: 2025-08-12T09:20:20.680Z

inefficiencies, inefficiencies everywhere

~ updated at: 2025-08-11T10:19:13.602Z

it’s a good world to live in when you engineer surprises to be good ones.

~ updated at: 2025-08-10T18:00:02.869Z

“the hype is how sama whips the members of technical staff into shape” ~ Roon

ref: https://x.com/tszzl/status/1953601056526807456

~ updated at: 2025-08-08T06:30:05.135Z

Notes from “In Support of Shitty Types” by Mitsuhiko

Go’s types are much less expressive and very structural. Things conform to interfaces purely by having certain methods. The LLM does not need to understand much to comprehend that. Also, the types that Go has are rather strictly enforced. If they are wrong, it won’t compile. Because Go has a much simpler type system that doesn’t support complicated constructs, it works much better—both for LLMs to understand the code they produce and for the LLM to understand real-world libraries you might give to an LLM.

one mustn’t forget that if interface{} and type-casting are used, they can still lead to runtime errors. This post primarily speaks about compile-time efficiencies.

Armin ends the post with a very interesting commentary:

I think it’s an interesting question whether this behavior of LLMs today will influence future language design. I don’t know if it will, but I think it gives a lot of credence to some of the decisions that led to languages like Go and Java. As critical as I have been in the past about their rather simple approaches to problems and having a design that maybe doesn’t hold developers in a particularly high regard, I now think that they actually are measurably in a very good spot. There is more elegance to their design than I gave it credit for.

I believe A.I. can be a lens to appreciate/depreciate decisions whose value that may not have been fully evident.

~ updated at: 2025-08-05T17:30:02.819Z

“you can either spend $100/mo on groceries or claude max and only one of those is going to make you a millionaire”

replace subjects and numbers with anything and that’s a good argument.

it brings out two things:

1. it takes away the focus from lower limit onto the upper limit

2. sprinkles specificity and subjectivity

arguments against don’t seem to hold for the person making the statement in a acute perspective.

ref: https://x.com/rhyssullivan/status/1951832934245433700?s=46

~ updated at: 2025-08-03T17:30:05.315Z

there are fewer known pleasures than listening to trinity while working weights.

~ updated at: 2025-08-02T17:30:07.591Z

Paul Graham writes all his essays in vim.

I think I already knew that but also didn’t.

ref: https://x.com/paulg/status/1951377078114460075?s=46

~ updated at: 2025-08-02T17:30:05.027Z

An excerpt from a rant by Salvatore Sanfilippo on X.

Companies discover that complexity is a form of vendor lock-in, among the other things. People that created a mess get promoted, since IT no longer evaluates design quality as one of the metrics for success.

ref: https://x.com/antirez/status/1950483014938484928

~ updated at: 2025-07-30T17:30:02.595Z

Excerpt from “Agentic Coding Things That Didn’t Work”

Forcing myself to evaluate the automation has another benefit: I’m less likely to just blindly assume it helps me.

Because there is a big hidden risk with automation through LLMs: it encourages mental disengagement. When you stop thinking like an engineer, quality drops, time gets wasted and you don’t understand and learn. LLMs are already bad enough as they are, but whenever I lean in on automation I notice that it becomes even easier to disengage. I tend to overestimate the agent’s capabilities with time. There are real dragons there!

You can still review things as they land, but it becomes increasingly harder to do so later. While LLMs are reducing the cost of refactoring, the cost doesn’t drop to zero, and regressions are common.

I too have experienced this “mental disengagement”. I think this traces back to my mindset of “DRY” (Don’t repeat yourself) and “Solve once, use forever”

It’s funny how literal my subconsicous takes these rules. I have to consicously remind myself that I don’t live in an ideal world where everything is deterministic.

REF: https://lucumr.pocoo.org/2025/7/30/things-that-didnt-work/ by @mitsuhiko

~ updated at: 2025-07-30T15:29:10.019Z

Terence Tao on applying Blue team vs Red team concepts to LLMs:

In the field of cybersecurity, a distinction is made between the “blue team” task of building a secure system, and the “red team” task of locating vulnerabilities in such systems. The blue team is more obviously necessary to create the desired product; but the red team is just as essential, given the damage that can result from deploying insecure systems.

The nature of these teams mirror each other; mathematicians would call them “dual”. The output of a blue team is only as strong as its weakest link: a security system that consists of a strong component and a weak component (e.g., a house with a securely locked door, but an open window) will be insecure (and in fact worse, because the strong component may convey a false sense of security). Dually, the contributions to a red team can often be additive: a red team report that contains both a serious vulnerability and a more trivial one is more useful than a report that only contains the serious issue, as it is valuable to have the blue team address both vulnerabilities. (But excessive low-quality reports can dilute attention from critical issues.)

Because of this, unreliable contributors may be more useful in the “red team” side of a project than the “blue team” side, though the blue team can still accommodate such contributors provided that the red team is competent enough to catch almost all of the errors that the contributor to the blue team might make. Also, unreliable red team contributions only add value if they augment the output of more reliable members of that team, rather than replace that output, and if their output can be effectively filtered or triaged by more experienced red team members. (1/3)

The blue team / red team distinction extends beyond cybersecurity to many other disciplines as well. In software engineering, for instance, “blue teaming” might correspond to the generation of new computer code, while “red teaming” would consist of such tasks as quality assurance and testing of such code. In mathematics, “blue teaming” could involve coming up with speculative ideas to solve a math problem, while “red teaming” checks the arguments for formal errors, and also raises heuristic objections to a blue team approach being viable. (See also my discussion about “local” and “global” errors in mathematics at https://terrytao.wordpress.com/advice-on-writing-papers/on-local-and-global-errors-in-mathematical-papers-and-how-to-detect-them/ ).

I like to refer to these two teams in mathematics as the “optimists” and “pessimists”; in my experience, the strongest collaborations arise when there is a roughly equal split between the optimists and pessimists in the collaborations. (Depending on the collaboration, I myself have sometimes played the optimist, sometimes the pessimist, and sometimes a mixture of both.) (2/3)

Many of the proposed use cases for AI tools try to place such tools in the “blue team” category, such as creating code, text, images, or mathematical arguments in some semi-automated or automated fashion, that is intended for use for some external application. However, in view of the unreliability and opacity of such tools, it may be better to put them to work on the “red team”, critiquing the output of blue team human experts but not directly replacing that output; “blue team” AI use should only be permitted up to the capability of one’s “red team” to catch and correct any errors generated. This approach not only plays to current AI strengths, such as breadth of exposure and fast feedback, but also mitigates the risks of deploying unverified AI output in high-stakes settings.

In my own personal experiments with AI, for instance, I have found it to be useful for providing additional feedback on some proposed text, argument, code, or slides that I have generated (including this current text). I might only agree with a fraction of the suggestions generated by the AI tool; but I find that there are still several useful comments made that I do agree with, and incorporate into my own output. This is a significantly less glamorous or intuitive use case for AI than the more commonly promoted “blue team” one of directly automating one’s own output, but one that I find adds much more reliable value. (3/3)

via Mastodon

~ updated at: 2025-07-30T10:18:04.780Z

Quoting Tim Sweeney on monitors

if you’re gonna spend 4500 hours per year in front of a monitor, buy the best.

ref: https://x.com/timsweeneyepic/status/1949224032827846954?s=46

~ updated at: 2025-07-27T17:30:04.983Z

Uncle Bob on SQL

SQL was never intended to be used by computer programs. It was a console language for printing reports. Embedding it into programs was one of the gravest errors of our industry.

ref: https://x.com/unclebobmartin/status/1917410469150597430

~ updated at: 2025-07-25T17:30:04.929Z

the only thing about A.I. prompting that bothers me is that I’m not thinking…

i’m not in flow state.

it’s like im constantly being interrupted by slop.

i don’t mind being wrong and retrying…

it’s always like here’s the requirement, here’s the function signature, now give full function. I just copy and paste, test, review, git commit.

it’s lovely for small bug fixes, not great for learning something.

i get why technical founders who are looking to increase customer-value per effort really love it.

using A.I. forces you to be an owner looking for value.

the denominator is falling exponentially. i need to change by mindset.

~ updated at: 2025-07-25T13:09:05.699Z

Quoting Paul Graham on starting a software startup:

If you want to start a software startup, you should still learn to program. Even if AI writes most of your code, you’ll still be in the position of an engineering manager, and to be a good engineering manager you have to be a programmer yourself.

But there’s another even more important reason to learn to program. That’s how you get ideas for startups: when you look around the world with a programmer’s eye, you see all the things that could be built but haven’t been yet.

If you have the ability to exercise agency, you will find a way to get things done. The easiest way to get things done is do it first in your head.

ref: https://x.com/paulg/status/1947639333214687700

~ updated at: 2025-07-22T13:39:22.071Z

it’s not agi if it starts every reply with “you’re absolutely right”

~ updated at: 2025-07-22T01:16:05.206Z

i really wish @x had a free tier API access to retrieve bookmarks.

~ updated at: 2025-07-20T17:30:02.646Z

advice to young kids by john carmack:

“knowledge and depth” “through and through” “find inefficiencies that can be bypassed” “its not necessary, be reliable, do quality work” “but people who do have inclination to doing things deeper, there are layers of things out there, it’s amazing” “it’s better than ever” “whole new worlds to explore” “hard work, understand as much as you can” “you will make more total progress, by telling yourself that there is a whole broad world around me” “preparing myself with broad tools and im being aware of change” “looking for opportunies to deploy the tools you’ve built” “go deeper and bypass inefficiencies” “make things better”

~ updated at: 2025-07-20T14:32:30.407Z

Salvatore Sanfilippo uses Gemini 2.5 Pro / Claude to fix vector set bugs in redis core.

Here are some excerpts that I found interesting:

The most famous LLMs are not the best. Coding activities should be performed mostly with:

• Gemini 2.5 PRO
• Claude Opus 4

Gemini 2.5 PRO is, in my experience, semantically more powerful. Can spot more complex bugs, reason about more complex problems. Claude Opus may be better at writing new code sometimes (sometimes not), the user interface is more pleasant, and in general you need at least two LLMs to do some back and forth for complex problems in order to enlarge your (human) understanding of the design space. If you can pick just one, go for Gemini 2.5 PRO.

I have been using only chatGPT whenever I use an LLM to help me get some coding/knowledge work done. Looks like I gotta start using Gemini 2.5 pro now.

Always be part of the loop by moving code by hand from your terminal to the LLM web interface: this guarantees that you follow every process. You are still the coder, but augmented.

I too prefer keeping my llm answers in my terminal or web browser itself. I don’t use A.I. in my IDE (vscode and sometimes neovim) because I like to keep control on what goes in my editor.

full post: https://antirez.com/news/154

~ updated at: 2025-07-20T13:52:11.441Z

back in my day, vibe coding meant copy-pasting code from stackoverflow.

~ updated at: 2025-07-20T09:40:39.377Z

I fed an LLM a journal entry asking it specific questions about different problems I’ve have been facing and after the long discussion, it told one thing that resonated with me:

“perfectionism is surgical and not blanket”

~ updated at: 2025-07-19T17:30:02.527Z

TIL: golang has a sort package that has sort string method.

and the strings will be sorted by alphabetical/lexicographical order.

Playground Instance Link

…an alternative way (and a much better way) is using the slices package and calling slices.Sort() which is agnostic of the data type.

~ updated at: 2025-07-19T11:44:06.798Z

I found this tweet from naval:

Technology will eventually have to drive human evolution (embryo selection, gene editing).

Otherwise, our technology will evolve faster than our biology and lead to Paleolithic humans facing Neolithic stressors.

it does link to the idea of the great filter and this statement popped into my head immediately after:

the great filter is just stressors in different costumes

~ updated at: 2025-07-19T04:28:28.687Z

i should 2x my agency and then lock in

~ updated at: 2025-07-18T03:06:20.307Z

today’s light workout

6s x 10r chest (fly ups & press) 3s x 10r biceps 20 minute cardio

~ updated at: 2025-07-15T04:26:45.615Z

TIL: that john carmack once used Anki [^1] to learn vim keybindings.

reference: https://youtu.be/tzr7hRXcwkw?si=Tm7dFf0d_nqILXb2&t=746 at 12:29 mark

[^1] https://apps.ankiweb.net

~ updated at: 2025-07-13T17:30:05.064Z

today’s light workout

10r x 6s Chest (fly ups & chest press) 10r x 3s Biceps 20 min Cardio 10 pushups

~ updated at: 2025-07-13T07:42:47.542Z

wrote a safari extension to block youtube shorts. installed on laptop and phone.

let’s see if my productivity goes up 🤞🏻

~ updated at: 2025-07-12T17:30:02.555Z

Jack starts permissionlesstech

i did this as a weekend project to understand a few technologies: bluetooth mesh, store and forward and relay models, and various encryption models. this was one part of a series of challenges i’ve given myself: to build something every day that i didn’t think i was capable of, and that i didn’t think the current set of AI tools were capable of, namely block’s tool goose (https://block.github.io/goose/) . i’m always surprised at how much…just works. we have a new way of programming now, with an entirely new language and compiler: english and intelligence models.

ref: https://github.com/orgs/permissionlesstech/discussions/139

~ updated at: 2025-07-11T03:34:20.910Z

Brian Armstrong on “Share a piece of lore about yourself”

He shared a struggle he faced during the early phases of Coinbase. It’s a long post, so I’m linking it here for reference .

He ends the story with this,

It’s a great testament to how constraints breed creativity, top talent matters in startups, and teams are often capable of more than they think when there is no other option. Most products that succeed have early moments like this, where someone has to step up and make a play on the field that defies all the odds. As we face new challenges and deadlines across our many products, I always look out for who on the team is ready to step up and make the game-winning play on the field.

I do recommend that you read his entire post.

~ updated at: 2025-07-09T03:25:54.394Z

“someone please…”

“…please…”

“…please review my code”

~ updated at: 2025-07-08T08:49:21.688Z

Watching a video is like skimming the material where by the end, you’ll know where it’s all headed and the mind is already working on recognizing the independent points of agency, a traversal in the graph if you will.

Then as a second pass, slowing re-reading the transcript to help the mind affirm/reaffirm or even break assumptions, is gonna have a lasting impact in the learning process.

This explains why I remember random lore/moments in stories I adore. Watch something end to end and then meticulously revisit parts with a new perspective (akin to multimodal recall).

Inspired by this wonderful piece of writing: https://cse.buffalo.edu/~rapaport/howtostudy.htm

~ updated at: 2025-07-06T17:30:09.778Z

“I know you think you understand what you thought I said, but I am not sure you realize that what you heard is not what I meant”

~ updated at: 2025-07-05T11:50:10.012Z

Watching a video is like skimming the material where by the end, you’ll know where it’s all headed and the mind is already working on recognizing the independent points of agency, a traversal in the graph if you will.

Then as a second pass, slowing re-reading the transcript to help the mind affirm/reaffirm or even break assumptions, is gonna have a lasting impact in the learning process.

This explains why I remember random lore/moments in stories I adore. Watch something end to end and then meticulously revisit parts with a new perspective (akin to multimodal recall).

Inspired by this wonderful piece of writing: https://cse.buffalo.edu/~rapaport/howtostudy.htm

~ updated at: 2025-07-05T11:49:23.200Z

I might have to experiment with daily setup and teardown rituals because mission critical work can be energy draining.

~ updated at: 2025-07-03T08:39:13.146Z

i wrote myself a neovim command and a keybinding on top of that to help me save and push the current file to remote git branch.

If I want to quickly save current file and push to remote all i need to do it

gp

This is super helpful when working with personal single file notes and scripts.

my full chatGPT conversation transcript

~ updated at: 2025-06-28T07:14:07.655Z

The Pen Is Mightier Than the Keyboard by Mueller and Oppenheimer found that students who took handwritten notes retained concepts better than those who typed them.

Taking notes in your own words along with verbatim (quotes) and can help later validate and rethink your assumptions.

Hand writing in your own words + digital / analog verbatim transcribing i feel is in line with the Lollapalooza¹

~ updated at: 2025-06-28T05:12:55.521Z

I wanted to improve my code review capability and I found this blog post useful!

so sharing it here just because: https://consulting.drmaciver.com/code-review-quick-fixes/

~ updated at: 2025-06-26T12:30:02.812Z

Rosetta 2 creator on efficiency:

Q(zmb_ ): […]As someone frustrated in a team of 10+ that is struggling to ship even seemingly trivial things due to processes and overheads and inefficiencies, I would really appreciate some insights on how do you organize the work to allow a single developer to achieve this.

A(cwzwarich): Well, the first thing to realize about scaling codebases with developers is that an N developer team will usually produce a codebase that requires N developers to maintain. So by starting small and staying small until you reach a certain critical mass of fundamental decisions, you can avoid some of the problems that you get from having too many developers too early. You can easily also fall into the reverse trap: a historical core with pieces that fit too well together, but most of the developers on the team don’t intuitively understand the reasons behind all of the past decisions (because they weren’t there when they happened). This can lead to poorly affixed additions to a system in response to new features or requirements. As far as Rosetta in particular was concerned, I think I was just in the right environment to consistently be in a flow state. I have had fleeting moments of depression upon the realization that I will probably never be this productive for an extended period of time ever again.

sourced via comments to the HN post Rosetta 2 creator leaves Apple to work on Lean full-time

~ updated at: 2025-06-26T03:23:52.279Z

I rewatched Manchester by the Sea last night.

Casey Affleck really did a phenomenal job portraying unease and depression but Michelle Williams stole every scene she was in.

Moreover what kept me fascinated was the portrayal of climate/weather metaphorically and literally carrying the emotional journey and plot respectively.

The latter which is something that stood out to me because I’ve been dabbling with such metaphors in a story I’ve been brewing in my head.

~ updated at: 2025-06-26T02:55:44.675Z

I’ve been doing this fun thing lately.

Whenever I recall a concept that is practically not relevant to what I’m doing in my daily life (you can call me a normie) and I wanna see how people are dealing with said concept in their daily lives, I do this simple trick.

I search site:news.ycombinator.com {{concept}} on google and honestly I’m not disappointed. Instant rabbit hole material.

all this given that I have time to kill :P

~ updated at: 2025-06-26T02:45:22.954Z

Excerpts from “The Shape of the Essay Field” https://paulgraham.com/field.html

So the three reasons readers might not already know what you tell them are (a) that it’s not important, (b) that they’re obtuse, or (c) that they’re inexperienced.

so by implication,

If you’re writing for smart people about important things, you’re writing for the young.

Now that I know it, should I change anything? I don’t think so. In fact seeing the shape of the field that writers work in has reminded me that I’m not optimizing for returns in it. I’m not trying to surprise readers of any particular age; I’m trying to surprise myself.

~ updated at: 2025-06-24T04:08:40.622Z

code that runs a million times needs thinking atleast twice.

~ updated at: 2025-06-23T06:57:22.757Z

a court order over trademark dispute from “iYO” against “iO” to take down an announcement page is crazy

“We don’t agree with the complaint and are reviewing our options.” ~ OpenAI

https://x.com/OpenAINewsroom/status/1936910167867863053

~ updated at: 2025-06-23T02:45:29.812Z

While trying to learn about setting keyboard shortcuts via vanilla javascript, I learned that navigator.platform showed MacIntel despite my device being a post 2020 Apple M2 Pro .

I was confused. I needed to know if I should be.

Turns out that I needn’t because there was this thing called Rosetta 2 that was intentionally returning MacIntel because it supported applications compiled to run on previously intended Intel X86 chips on now Apple owned silicon stack.

What is Rosetta 2? My notes: https://tnvmadhav.me/til/rosetta-2/

~ updated at: 2025-06-22T08:42:25.411Z

T.I.L. that A.R.M. initially stood for Acorn R.I.S.C. Machine and was rebranded to A.R.M. which stands for Advanced R.I.S.C. Machines Ltd.

~ updated at: 2025-06-22T08:02:47.409Z

T.I.L. that PowerPC was an alliance between Apple, Motorola and I.B.M.

~ updated at: 2025-06-22T07:59:48.395Z

I was almost certain that @digitalocean had a way to configure event alerts using webhooks.

after a clunky search, any attempt to land on an opportunity on closing this requirement led me to a haven of expectations or just ideas dot digitalocean dot com.

like what even?

The aforementioned haven of expectations: https://ideas.digitalocean.com/core-compute-platform/p/add-webhooks-to-monitoring

~ updated at: 2025-06-22T07:07:29.289Z

Content I consumed today

1. Software is changing again [Video] https://www.youtube.com/watch?v=LCEmiRjPEtQ

2.Future of AI [Video] https://youtu.be/mZUG0pr5hBo (weird and wholesome to hear sama without his signature vocal fry)

~ updated at: 2025-06-19T17:30:02.614Z

Content I consumed today:

1. Writing Toy Software Is A Joy https://www.jsbarretto.com/blog/software-is-joy/ (good arguments)

2. The Grug Brained Developer https://grugbrain.dev (funny, relatable)

3. “Don’t Mock What You Don’t Own” in 5 Minutes https://hynek.me/articles/what-to-mock-in-5-mins/ (obvious in hindsight)

~ updated at: 2025-06-19T03:44:01.710Z

Running notes from Software is changing (again)

software 1.0 -> computer code software 2.0 -> language model weights

Software 2.0 is eating Software 1.0

A lot of software 1.0 or even 2.0 is being rewritten.

“LLMs are people spirits”

a time machine gone wrong (if you will)

LLMs are like a person when their character-arc is complete.

LLMs generate while humans need to verify (as of today)

Going forward we gotta (as humans) setup systems that are intuitive, easy to modify and use. Fast recognition of LLM induced BS.

Karpathy’s Slide:

1. describe single, next concrete, >incremental change
2. don’t ask for code, ask for approaches
3. test
4. git commit
5. ask for suggestion on what could be implemented next
6. repeat

vague prompts implies vague verification bias. spend time strengthening verification. prompts get stronger as a result.

So test driven development with more steps but when done right, we can get a lot more done. many people will not like this and they’d spend money 💰 to get it done for them. good business opportunity.

however being money minded without having the tenacity to solve the problem will lead to blow up. consumers of such businesses/solutions should decide before jumping in on the hype.

~ updated at: 2025-06-19T03:42:21.003Z

what i can’t fix i don’t understand https://twitter.com/twitter/status/1935170294467211626/

~ updated at: 2025-06-19T02:46:47.967Z

Content I consumed today:

1. Writing Toy Software Is A Joy https://www.jsbarretto.com/blog/software-is-joy/ (good arguments)

2. The Grug Brained Developer https://grugbrain.dev (funny, relatable)

3 “Don’t Mock What You Don’t Own” in 5 Minutes https://hynek.me/articles/what-to-mock-in-5-mins/ (obvious in hindsight)

~ updated at: 2025-06-18T16:08:01.250Z

I came across a website while surfing hacker news today.

https://theoremoftheday.org

Daily dose of theorems. Neat :)

~ updated at: 2025-06-18T12:30:02.826Z

“Don’t Mock What You Don’t Own” in 5 Minutes https://hynek.me/articles/what-to-mock-in-5-mins/

Some points that caught my attention:

To keep your business code testable and idiomatic, avoid directly using third-party dependencies in it.

Business logic is ultimately the reason you write software. It’s the reason an application exists. Having clean and idiomatic business logic pays dividends for as long as the software exists which is always longer than you think.

~ updated at: 2025-06-18T10:07:04.704Z

Writing Toy Software Is A Joy https://www.jsbarretto.com/blog/software-is-joy/

Some sentences that caught my attention:

“What I cannot create, I do not understand” “[…]the joy in building toy projects like this comes from an exploration of the unknown, without polluting one’s mind with an existing solution.[…]”

“I’ve been consistently surprised by just how often some arcane nugget of knowledge I’ve acquired when working on a toy project has turned out to be immensely valuable in my day job, either by giving me a head-start on tracking down a problem in a tool or library, or by recognising mistakes before they’re made.”

“If you’ve been using LLMs for a while, this cold-turkey approach might even be painful at first, but persist. There is no joy without pain.” “The runner’s high doesn’t come to those that take the bus.”

~ updated at: 2025-06-18T03:01:30.620Z

On the foolishness of nature programming by Dijkstra

https://www.cs.utexas.edu/~EWD/transcriptions/EWD06xx/EWD667.html

~ updated at: 2025-06-11T14:02:22.010Z

making a dog vibe code an app should be something someone should attempt.

transcribe dog barking, ask chatGPT to translate to english with a context of B2B SaaS, ask cursor to process it, build it.

~ updated at: 2025-06-08T08:18:19.725Z

I stumbled upon a post from 2010 with a very interesting title. https://www.lesswrong.com/posts/ufBYjpi9gK6uvtkh5/for-progress-to-be-by-accumulation-and-not-by-random-walk

I agree with the fact that top down learning(intentional progress) is much more efficient than bottom up learning (random walks).

I believe this was something I supported for a long time now.

~ updated at: 2025-06-08T07:11:47.543Z

the type of person that listens to 10+ minutes of lexfridman ad. reads just to feel something

~ updated at: 2025-06-06T07:11:50.575Z

Little late to the party and I just finished watching Mindhunter season 2.

OMG just a fascinating gripping show.

it’s a shame netflix canceled season 3. come on @netflix you can still make it happen.

~ updated at: 2025-05-30T15:56:29.365Z

Tim Sweeney says that we gotta do a lot of different types of work to gain maturity and like don’t focus on only one type.

do a lot of things do a lot of coding a lot of practicing a lot of different styles and only then you will know the true sense of coding and then you can choose to go deep once you’ve gain mastery in all different strokes

This is super interesting coming from a hardcore engineer where specialisation seems to be the thing but craft is honed though creativity and exploration.

I’ve been listening to the epic games CEO on lexfridman, i knew that he had been against apple’s store monopoly but i did not know about his background.

I loved fortnite during its peak 2019 i think, played almost daily with friends and strangers around the internet. smooth gameplay. dont miss being headshot by noobs though.

~ updated at: 2025-05-30T12:49:44.982Z

I just got jump-scared by @GitHub.

(all I did was merge a PR to main branch)

This seems to be a bug or something in the GitHub news feed on org dashboard.

~ updated at: 2025-05-30T06:52:09.270Z

Today I learned about tee terminal command.

~ updated at: 2025-05-29T09:53:33.126Z

GitHub is down (partially)

~ updated at: 2025-05-26T09:35:59.669Z

doing something and not knowing why so is not intentional enough. either find out why or do something else.

~ updated at: 2025-05-26T03:57:55.012Z

daniel pemberton’s OST composition for guy ritchie’s king arthur movie is underrated. it’s a shame that it didn’t do well in the box office.

barring the janky boss fight, there are some really powerful moments in the movie.

~ updated at: 2025-05-25T12:22:37.183Z

optimizing for easy addition/updation/removal of integration tests

~ updated at: 2025-05-20T06:08:28.693Z

mustering the courage to write integration tests (it’s too time consuming)

~ updated at: 2025-05-19T11:20:43.930Z

optimising for not getting nerd sniped

~ updated at: 2025-05-19T08:12:14.234Z

the tendency to optimize for the wrong loss function will eventually go belly up.

~ updated at: 2025-05-19T06:20:35.472Z

if anything, llms have helped us appreciate authentic human thoughts and ideas. long way to go.

~ updated at: 2025-05-18T13:22:08.682Z

bias is a hurdle to unlearning. bias is a hurdle. bias is overfitting.

~ updated at: 2025-05-17T07:45:44.840Z

relaxing the learning muscle is a superposition problem and it needs to stop being one.

will you get back to it or will you let it go? thankfully you have agency to make this choice.

~ updated at: 2025-05-17T07:36:18.128Z

I recently stumbled upon a nice explainer of the dutch flag sorting algorithm.

via YT search feature

With this you can sort three things in one pass without using extra space.

Next time I’ll use this to sort my sock drawer (if I get the time) 😂

~ updated at: 2025-05-17T04:17:21.619Z

With inaction, entropic ruin is imminent.

~ updated at: 2025-05-16T11:27:37.151Z

you are the moat

~ updated at: 2025-05-15T11:35:39.648Z

🎶 Not me listening to Zero Eclipse instrumental by Hiroyuki Sawano on repeat

~ updated at: 2025-05-15T03:28:14.261Z

I should maintain a running notes of blogs, articles and stuff that I read.

~ updated at: 2025-05-14T14:25:30.693Z

the type of person to do things that don’t scale.

~ updated at: 2025-05-14T14:07:45.166Z

Even with technology, the Darwinian natural selection is coming into play. https://twitter.com/twitter/status/1921687346946179419/

~ updated at: 2025-05-12T02:21:06.098Z

eren jaeger was fighting an internal struggle within himself since the infamous medal ceremony oscillating between “There is no free will” and “I did all of this with my free will”.

he couldn’t take it. he needed help from Armin to digest and reconcile that he is a “slave to freedom”.

and, this was triggered by that flash back to grisha holding new born eren in his arms saying “You are free” is such a powerful cathartic moment.

yes cathartic moments in pieces of fiction is my weakness.

🏞️ Context #1 , 🏞️ Context #2 , 🏞️ Context #3 , 🏞️ Context #4 , 🏞️ Context #5

~ updated at: 2025-05-11T16:28:02.561Z

true understanding is about knowing and confidently predicting interaction and not just isolated behaviour. https://twitter.com/twitter/status/1921119501846016115/

~ updated at: 2025-05-10T08:31:06.376Z

I’m beginning to believe that a good learning process (leading to proper understanding) will emerge through a bottom up and top down encounters serendipitously.

akin to a pincer operation.

~ updated at: 2025-05-10T08:26:07.268Z

twitch dot tv but private, within the org.

this has potential to be fun, keep yourself and team accountable and be productive.

i believe there will be atleast 100 orgs that will be onboard this idea.

i didn’t google if there is already an existing solution or product for this. if yes it’s not popular enough.

~ updated at: 2025-05-10T05:11:50.010Z

something about unyielding primal rage in works of fiction that gives me goosebumps.

🏞️ Context #1

~ updated at: 2025-05-09T17:20:07.247Z

TIL that you can initiate GitHub action workflow from local using gh client as opposed to clicking through cluncky buttons for manual runs.

This is such a productivity boost for me.

gh workflow run deploy.yml --ref main

Appalled that I didn’t look for alternatives like this sooner.

~ updated at: 2025-05-08T10:04:57.036Z

“timelines are determined by members of technical staff” should be an axiom.

~ updated at: 2025-05-07T02:46:09.894Z

finally sat down and solved a icky bug in my content publishing software. practiced by debugging muscle on a relatively legacy piece of software. felt good. thanks ipdb.

~ updated at: 2025-05-03T12:33:13.668Z

I “vibe coded” a @github action to archive my daily live feed as markdown blog posts.

This was initially done using a manually scaffolded LaunchDaemon plist on my macbook but the issue was if the device was offline (without internet connection) at the time of archiving, it would sometimes not run at all.

Debugging why plist cron jobs couldn’t run was something I didn’t want to do so I now delegated it to github actions instead.

~ updated at: 2025-05-03T11:43:25.324Z

Quoting comment on the License change of Redis from SSPL to AGPL

“I have a great deal of respect for antirez and rec[o]gnize him as a kind and benevolent member of the FOSS community, but no matter what Redis, Inc. announced or does, they have lost my trust for good, and I will continue to use Redis forks for as long as they exist.”

via HN

loss of trust is a one way door . Once you cross it, in most situations it’s hard to get back and needs incredible effort and sacrifice to gain back lost trust.

Integrity on the fly is important. I know that things won’t be smooth sailing always and the absurdist in us agrees that we may have to take decisions that we ideally won’t.

that is exactly why decision makers need to have skin in the game on the long run.

~ updated at: 2025-05-02T04:49:52.220Z

I really hope vibe coding stops after going from zero to one.

The term vibe engineering sends shivers down my spine but one man’s fear is another man’s opportunity.

When push comes to shove, we must be able to turn a vibe coded product to a well engineered sound product and business.

…or am I being myopic about all this?

~ updated at: 2025-04-28T01:37:09.021Z

overthinking is anticipating the nth order effect of an action or lack thereof.

overthinking often collapses to a delusion. then, emotional turmoil.

instead, how about we try to stack actions that seem to yield only positive growth regardless of feedbacks leading to different nth order effects?

then, we’re just indirectly limiting decisions to 1st order effects only make things easier on the mind. This curtails overthinking & its negative effects by several orders.

Not speaking for positive effects of overthinking cuz it’s not productive 😅 for me.

~ updated at: 2025-04-25T08:29:05.399Z

For the long term, it’s important to know what is actually going on because a project is subject to modifications and extensions.

Speaking of extensions and modifications, vibe coding on the other hand calls for complete rewrite and, though A.I could eventually do a complete rewrite to support further updates. It feels we’re disrespecting the G.P.U.s and resources to get things done in a suboptimal way (if you will).

https://x.com/karpathy/status/1915581920022585597

~ updated at: 2025-04-25T03:33:43.901Z

intellectually stimulating problems to solve rather than burdens to bear

~ updated at: 2025-04-22T05:55:14.405Z

I got caught off guard LMAO 🤣

🏞️ Context #1

~ updated at: 2025-04-20T17:20:07.554Z

I came across a post that sheds light on the origin of VIM: https://pikuma.com/blog/origins-of-vim-text-editor

~ updated at: 2025-04-19T10:36:41.816Z

an A.I. fine-tuned to make your pool of engineers efficient and faster every day.

~ updated at: 2025-04-18T12:10:16.812Z

scrolling social media is like incentivising perpetual ephemeral thoughts and ideas.

ideally one should gain something out of it but we ignore going deeper into certain interesting thoughts and instead, keep surfing 🏄

that is not why we should use social media.

posters are getting something done and posting about it but consumers are just window-shopping (weird way to put it)

~ updated at: 2025-04-18T04:28:15.490Z

I’m strangely attracted to the idea of commit driven development.

Break down the task into list of commits annotated by the commit messages in great detail.

Tackle each of them in sequence.

It helps me keep stick to commit message while doing the changes.

It’s like a todo list but specific to low level implementation details.

~ updated at: 2025-04-17T07:15:58.092Z

maybe my job at the end of the day is to decrease entropy.

~ updated at: 2025-04-14T11:58:36.192Z

Make moves to bring about positive change instead of proving / disproving existing data points about results of the moves pool (not a pool as it’s unbounded)

~ updated at: 2025-04-09T03:36:32.146Z

I came across a post by @DHH about hiring developers¹. I opened it and this piece hit hard:

Maybe it sounds a little harsh, but a programmer who’s been working professionally for five years has likely already revealed their potential. What you’re going to get is roughly what you see. That doesn’t mean that people can’t get better after that, but it means that the trajectory by which they improve has already been plotted.

~ updated at: 2025-04-09T03:28:56.281Z

“low level execution punctuated by high level calibration”

~ updated at: 2025-04-09T03:24:46.553Z

it suddenly struck me that sustained self vigilance stops becoming a battle with the elimination of self imposed, limiting lens and that trigger to change isn’t always a moment in time but can be (and is initially) relentless sustained vigilance across several contiguous moments in time.

~ updated at: 2025-04-06T08:42:39.893Z

working on the slopes instead of the y-intercepts is the secret to optimism.

~ updated at: 2025-04-06T08:19:57.893Z

I witnessed a guy build a functional piano in his FPS game.

This dude is a legend. You have to leave being inspired.

https://www.youtube.com/live/ys5hays0TYo

~ updated at: 2025-04-06T04:28:01.471Z

whatever the equivalent of this is in your field

🏞️ Context #1

~ updated at: 2025-04-04T06:30:05.784Z

notes to self based on what i’ve observed:

if you think you’re slipping up often, try to spend more time and energy staying as low level as possible.

low level execution punctuated by high level calibration.

~ updated at: 2025-03-31T04:20:01.843Z

I came across an impressive thing while sipping my morning coffee.

A dude behind a twitter account (@PJaccetturo) generated the ‘Lord of the Rings: Fellowship of the Ring’ trailer in the style of Studio Ghibli (and it’s pretty impressive to reiterate).

The A.I. Filmmaker shared how the trailer was made.

My process was: Screenshot all 102 shots in the trailer Remix them to Ghibli style in @OpenAI’s Sora Animate in @Kling_ai and @LumaLabsAI

Re-edit in sync with trailer.

I’m attaching a hyperlink to the tweet that contains the generated trailer: https://x.com/PJaccetturo/status/1905151190872309907

Right now, it’s possible to do incredible stuff by duct-taping a bunch of tools together. I only see this method getting easier and easier as more people do nice and wholesome things with them.

~ updated at: 2025-03-28T03:37:56.858Z

the only way around a problem is through it.

~ updated at: 2025-03-27T05:59:43.108Z

so much to do, so little time…

~ updated at: 2025-03-27T03:11:29.824Z

The cost of doing code reviews is the obligatory burden of the whole context loaded in your brain before you start reviewing.

I try to take it as a self enforceable rule. It’s the least you can do.

It’s crazy that this has to be explicit.

~ updated at: 2025-03-25T09:35:22.557Z

While skimming through Simon’s list of posts¹, I realised that I’ve found the term that I’d been looking for – “Semantic Diffusion” (where the original idea/meaning dampens with time/loose adoption)

Semantic diffusion occurs when you have a word that is coined by a person or group, often with a pretty good definition, but then gets spread through the wider community in a way that weakens that definition. This weakening risks losing the definition entirely - and with it any usefulness to the term.

~ updated at: 2025-03-25T07:25:35.323Z

I stumbled upon (by the magic of the internet) this piece on Agency.

High Agency in 30 Minutes

One can use it as a compass 🧭 (perspective wise) while solving hard problems / making decisions.

~ updated at: 2025-03-25T07:22:59.751Z

Suhail shares his playbook on building A.I. powered products and services:

https://twitter.com/twitter/status/1903812003757736086/

~ updated at: 2025-03-23T15:05:09.438Z

that feeling when two orgs. you like because (they’re technically awesome) start throwing sh*t at each other

🏞️ Context #1 , 🏞️ Context #2

~ updated at: 2025-03-23T06:19:32.201Z

I’ve always had an appreciation to teams that ship products and features at an incredible pace. I’ve always thought it’s possible to do it without making mistakes. Still do.

~ updated at: 2025-03-23T04:14:47.888Z

The new research on using F.F.T. based optimization could help in making L.L.M.s faster, cheaper and able to handle longer contexts.

Linking the arxiv paper here: https://bagel.ink/c/kBT5Uw

~ updated at: 2025-03-22T16:06:20.417Z

linking original FFT paper.

An Algorithm for the Machine Calculation of Complex Fourier Series: https://bagel.ink/c/74YPp0

~ updated at: 2025-03-22T15:36:08.316Z

optimizing for cathartic experiences

~ updated at: 2025-03-09T13:01:37.731Z

Writing tests to validate your understanding of a piece of code should be something I must do more.

~ updated at: 2025-03-08T04:07:59.254Z

I’m seeing an increasing amount of x posts as ads. These are just random posts that don’t seem to sell or influence actively. These are just (seemingly) random posts about random stuff. What is the goal? gain a following through relatability? bizarre.

~ updated at: 2025-02-23T16:36:09.957Z

TIL what a .har file is (and how to generate them) https://tnvmadhav.me/til/http-archiving/

~ updated at: 2025-02-23T12:30:02.857Z

“If you think like a noob, you’re a noob” - GeoHotz

good advice.

~ updated at: 2025-02-23T09:11:15.933Z

just when i thought the x algorithm was doing a good job for me, it started to raid my timeline with posts from one account that I might’ve spent a couple of more seconds on than I should’ve.

(it was a robotics based post)

~ updated at: 2025-02-23T04:33:08.657Z

Really good blog post by Harper Reed: https://harper.blog/2025/02/16/my-llm-codegen-workflow-atm/

~ updated at: 2025-02-22T12:30:03.071Z

TIL about an in-built browser feature in V.S. Code (and likely why I didn’t know about it sooner) https://tnvmadhav.me/til/vscode-has-a-simple-browser/

~ updated at: 2025-02-22T09:39:28.519Z

I’d like to vouch for a tool that has contributed to my significant productivity boost (as a developer).

I’ve been using the lazygit C.L.I. tool on my machine and with a certain hotkey combo, the whole standard GIT ops (to ship changes) has been pleasurable.

~ updated at: 2025-02-22T09:36:07.369Z

“If you think like a noob, you’re a noob” - GeoHotz

good advice.

~ updated at: 2025-02-22T04:59:48.613Z

I should read the SOCK protocol https://www.rfc-editor.org/rfc/rfc1928

🏞️ Context #1

~ updated at: 2025-02-21T08:30:22.221Z

🎧😌🎶 Listening to ‘Light no Theme’ by Hideki Taniuchi

Simple yet a banger.

~ updated at: 2025-02-10T13:55:37.501Z

Generative A.I. models are a result of capturing & reproducing expertise from a moment in time.

Instead of vertices, the edges’ scaler properties are replicated with the statistical preface.

So zooming out, this could be viewed as / an apparent attempt to recreate a time machine gone wrong, or just having reached a local maximum, if you will.

~ updated at: 2025-02-10T03:00:24.572Z

“An idea like a ghost, must be spoken to a little before it will explain itself” - Charles Dickens

~ updated at: 2025-02-08T12:30:03.057Z

🎧🧑‍💻 Listening to ‘Deep Dive into LLMs like ChatGPT’ by Andrej Karpathy¹

~ updated at: 2025-02-08T07:46:01.447Z

🎧😌🧘🎶 listening to ‘To Give a Marionette Life’ by Yoshihisa Hirano

~ updated at: 2025-02-06T12:04:31.477Z

asking yourself “what would carmack do?”

🏞️ Context #1

~ updated at: 2025-02-01T16:41:15.772Z

But to increase collective IQ (rather than collective output), there needs to be something else.

— Think about problems differently than they’re used to

— For people who want to grind and solve problems in a better way, you now have a infinitely patient copilot/agent to help you complete the journey that you couldn’t complete before

— So, Intrinsic motivation is also necessary

— People need to start believing in increased individual agency and gain hunger to do things better

— Seems like we’re gonna be stuck in a local maxima otherwise https://twitter.com/twitter/status/1885560914000437650/

~ updated at: 2025-02-01T05:37:26.664Z

plain text only interaction for only those who prefer it.

For full utilisation, app layer innovation is the key (to support a larger part of market). https://twitter.com/twitter/status/1885179324224925752/

~ updated at: 2025-02-01T05:28:50.825Z

so much to do so little time…

~ updated at: 2025-01-31T07:06:42.798Z

List of phrases from lives of high agency people:

you can just do things you need to be more delusional it’s time to build don’t ask for permission move fast and break things the quick shall inherit the earth live players go direct founder mode man in the arena don’t die ¹ —- does it defy the laws of physics? there’s no unsolvable problem everything is a skill issue adults don’t exist there’s no way all normal behaviour is forgotten. only weird behaviour survives. one giant game of Roy optimise for the best story the amygdala is outdated hardware experiments > decisions end of day > end of week what have you got done this week? what is ignored by the media that will be studied by historians? questions are the answers you might need have you tried just doing the thing 100 times? specific ambition gives direction. general ambition gives anxiety. modern schooling is the low agency industrial complex everyone too busy worried what u think of them to notice u ²

~ updated at: 2025-01-30T13:19:35.632Z

“open source everything” ~ Jack, via X

~ updated at: 2025-01-28T15:00:03.002Z

Genuinely think the race to A.G.I. needs the rivalry of this ilk and yes it has begun.

https://x.com/LiangWenfeng_/status/1884080875450818754

Ok this account is not the official wenfeng lmao

~ updated at: 2025-01-28T14:03:42.890Z

listening to marc andreessen on lex fridman…

https://x.com/TnvMadhav/status/1862128639531499725

~ updated at: 2025-01-28T13:30:02.882Z

“…and oh yes almost forgot that deepseek was a side project”

https://x.com/karpathy/status/1872362712958906460

~ updated at: 2025-01-27T14:00:24.691Z

me waiting for filmbuff to upload that one much awaited review and analysis episode.

🏞️ Context #1

~ updated at: 2025-01-26T04:01:46.221Z

Cal Newport on TikTok¹

When I sign in for the first time, TikTok asks me to choose my interests from a long list illustrated by cheerful emojis. I select “Life Hacks,” “Science and Education,” and “Sports.” Then I’m off. The first video shows the Clemson University baseball team playing an exhibition game against the Savannah Bananas, a professional touring squad. The Clemson infield, for some inscrutable reason, starts dancing. I swipe up. A new video begins, showing someone selecting shoes at a store. The video is only ten seconds long; by the time I’ve finished jotting down some notes, it has already started replaying. I hastily swipe again. The next video plays tranquil music while a car slowly drives toward Yosemite National Park. The algorithm must have noticed that I lingered on the wintry scene: the next video shows someone sweeping snow off a porch with some kind of rotating broom contraption. Then the feed takes a darker turn, which makes me want to scroll even faster. I see a news story about a person being pushed onto subway tracks in Manhattan—swipe—a Trump video set to ominous music—swipe—“Top 15 Most Ghetto High Schools in New Jersey”—swipe—and someone making fun of a server’s accent in a restaurant. I shut down the app.

But if he were to again, I really hope he doesn’t come across “brain rot” ever but a part of me wants him to go down that rabbit hole (and write about how he felt).

~ updated at: 2025-01-25T05:09:48.650Z

majority of dopamine rush in software engineering comes from being able to reproduce bugs that couldn’t be easily before.

~ updated at: 2025-01-24T11:43:23.105Z

“LLMs are awesome and limited”

LLMs are great at things that require relatively lesser effort than generating the input itself¹.

~ updated at: 2025-01-22T11:16:57.395Z

I found this super interesting HN thread¹ of instances/spaces where people are using small language models to solve specific problems

~ updated at: 2025-01-22T02:43:50.560Z

Here’s a curated list of software stories about crazy bugs.

Source: Software Folklore

🏞️ Context #1

~ updated at: 2025-01-17T04:01:06.208Z

So much to do so little time and energy.

~ updated at: 2025-01-15T13:58:42.539Z

The ‘Grok please explain this tweet’ button is such as good feature.

nice work @X

~ updated at: 2025-01-11T04:27:07.315Z

A nice post¹ on dealing with large codebases in general:

Summary

• Large codebases are worth working in because they usually pay your salary
• By far the most important thing is consistency
• Never start a feature without first researching prior art in the codebase
• If you don’t follow existing patterns, you better have a very good reason for it
• Understand the production footprint of the codebase
• Don’t expect to be able to test every case - instead, rely on monitoring
• Remove code any chance you get, but be very careful about it
• Make it as easy as possible for domain experts to catch your mistakes

~ updated at: 2025-01-08T12:30:03.744Z

The state of AI assisted programming by Gergely Orosz and Addy Osmani

https://open.substack.com/pub/pragmaticengineer/p/how-ai-will-change-software-engineering

~ updated at: 2025-01-08T07:22:39.151Z

Excerpts from Reflections (2025), a post by Sam Altman

“We are now confident we know how to build AGI as we have traditionally understood it. We believe that, in 2025, we may see the first AI agents “join the workforce” and materially change the output of companies. We continue to believe that iteratively putting great tools in the hands of people leads to great, broadly-distributed outcomes.”

" Superintelligent tools could massively accelerate scientific discovery and innovation well beyond what we are capable of doing on our own, and in turn massively increase abundance and prosperity."

Two years after incredible growth, OpenAI is moving the goalpost to A.S.I. beyond the relatively myopic A.G.I. goalpost[^1].

What sort of scaling laws apply here?

If A.S.I. is the goal, then is it okay for a “product” company to be the one to lead it’s efforts? No clear answers but the mind is curious.

[^1] I’m curious to see what sort of approach one(by that I mean an entire organization of industry-leading teams of researchers) would take to move to A.S.I.

~ updated at: 2025-01-06T01:54:06.149Z

Pacific Rim (2013) is such a good movie.

~ updated at: 2025-01-05T13:25:28.373Z

An A.I. agent that is trained on your taste. You tell it what you want (regularly or just ad-hoc), and give it constraints if you’re in a hurry. It scours the web, finds links, measures as per your taste, and gives you a compilation. You should be able to let it know what taste you’d like to develop. Inherit this agent and deploy it to learning, shopping, exploring, etc.

In short, it is all aspects of life that you want to grow in.

~ updated at: 2025-01-05T04:04:00.072Z

I really don’t know why but this snippet reminds me of the phrase

“[…]an unstoppable force meets an immovable object”

Not because I think there is a 1:1 mapping to this phrase but reminiscence.

Source

more for context

🏞️ Context #1

~ updated at: 2025-01-05T03:03:36.929Z

If you plan on levelling up your technical writing / blog post (meant for education), you’d like to visit this guide once in a while.

🏞️ Context #1

~ updated at: 2025-01-04T06:30:18.727Z

Yet again reminded of this.

🏞️ Context #1

~ updated at: 2025-01-04T06:30:07.062Z

every time I see Adam interacting with a nakamoto post, I’m reminded that he could most definitely be nakamoto 😆

~ updated at: 2025-01-04T04:28:16.127Z

Yet again reminded of this.

🏞️ Context #1

~ updated at: 2025-01-03T15:58:07.497Z

How to learn marketing and sales as a solo entrepreneur?

BitBasher

“I built and run a saas product solo (> $500k/year). Honestly, I have never found any advice useful[…]

[…]My business is b2b and it’s a platform for certain kinds of professionals as well as an API that powers many well known businesses.

1. Networking
2. Long Game w/ SEO, Word of Mouth

“I’d say focus on the long game from day one (blog posts, good marketing pages, etc).”

“If no one wants to talk to you about it, no one is going to want to pay for it.”

More… A bunch of resources linked in replies

~ updated at: 2025-01-03T15:55:05.498Z

If you plan on levelling up your technical writing / blog post (meant for education), you’d like to visit this guide once in a while.

🏞️ Context #1

~ updated at: 2025-01-03T12:34:54.244Z

I found myself reading the (well-written) Rules of Software Tutorials ¹.

I’d started publishing guides for beginners² and I found some value in the aforementioned post, that I will be inculcating in my future guides.

I’d highly recommend a read if you’re a developer planning to write guides or technical documentation.

~ updated at: 2025-01-03T12:20:17.146Z

I think it’s about time where we need be able to build stuff incredibly quickly like physically build infrastructure, incredibly quickly with technology.

~ updated at: 2025-01-03T07:00:03.320Z

it’s funny that it has to be said this way, but programming today is being mainly used to apply solved problems and build stuff on top of that but believe programming was introduced for different reason, to find solutions to problems that a human brain cannot explore, naturally with ease.

~ updated at: 2025-01-03T06:46:27.806Z

“One of my papers got declined today by the journal[…]This paper was also submitted elsewhere, and accepted” – Terence Tao, via Mastodon

[…]Because of this, a perception can be created that all of one’s peers are achieving either success or controversy, with one’s own personal career ending up becoming the only known source of examples of “mundane” failure. I speculate that this may be a contributor to the “impostor syndrome” that is prevalent in this field (though, again, not widely disseminated, due to the aforementioned reporting bias, and perhaps also due to some stigma regarding the topic).

Rejection is actually a relatively common occurrence for me, happening once or twice a year on average.

I find it best not to take these sorts of rejections personally, and move on to other journals, of course after revising the paper to address any issues brought up by the rejection.

You only fail if you give up.

~ updated at: 2025-01-02T06:30:05.896Z

John Carmack on LLMs and Interfaces:

LLM assistants are going to be a good forcing function to make sure all app features are accessible from a textual interface as well as a gui. Yes, a strong enough AI can drive a gui, but it makes so much more sense to just make the gui a wrapper around a command line interface that an LLM can talk to directly. ~ John Carmack

“The question is will top AIs get better at gui faster than all apps add text. I think I have a guess” - Andrej Karpathy

“There will probably be more conventional software written in the coming decade than ever before; the architecture of that software still matters! Better and worse affordances for AI should make a difference.” - John Carmack

When developing software, we tend to generally first try to build a Linux program that works on the terminal via CLI command, etc, with the intention of running unit tests like preliminary runs to satisfy the human mind.

Then, we integrate the said feature into whatever the product is.

This thread reminds me of Stripe’s DEV console CLI experience where every action that is possible via GUI can also be done via the Stripe CLI tool.

So building tools and apps must be looked at as layers/ levels where level 1 is a functional CLI app and level 2 is a GUI that is catered to a human being and their ease of use.

🏞️ Context #1

~ updated at: 2025-01-01T16:44:42.771Z

Guillermo Rauch, CEO of Vercel published a DOOM game-based captcha.

Kill 3 enemies to pass the captcha.

Surprisingly difficult if you move away from the spawn point.

doom-captcha.vercel.app

There’s a “How it works” section but there doesn’t seem to be a hard link to this section because it’s all controlled by Javascript (spectacular RauchG fashion :wink: :wink:) but here’s the source code

Here’s an X post by RauchG https://x.com/rauchg/status/1874130110120706556

~ updated at: 2025-01-01T16:15:32.679Z

I just stumbled upon a short and sweet post via Hacker News

It’s essentially a note about getting started with something and a mindset to keep going forward.

This post served as a good reminder to envision growth in the next quarter of the 21st century.

A few points inspired from the post:

1. Start small because expecting a big change will be overwhelming to the pondering mind

2. It’s okay even if you try to copy others because unconsciously and fortunately, you’ll eventually add your spin onto it, and with the aid of time, whatever you’re nurturing will turn out into something unique

3. What looks like overnight success for others, is just an insane amount of work and discipline put behind the scene

4. The effort someone who is at a stage you’re striving towards has put will be more or less in the same ballpark as how much you will have to put so no point in comparing yourself to others

Sometimes magic is just someone spending more time on something than anyone else might reasonably expect.

~ updated at: 2025-01-01T15:59:38.001Z

I just realized that the century is 25% done and a lot has happened.

Here’s to the next and stronger 25%!

~ updated at: 2025-01-01T13:54:07.343Z

“there’s an ancient class war at play. the culture of the euro/anglo aristocracy is of elegant leisure, art patronage, athletics. the culture of the capitalist bourgeoisie is of constant striving, sheer Faustian drive, efficiency”

“the former is much more tasteful but the latter pays the civilizational dues”

— Roon

There are times when every other role comes into play with increased ROI (civilization scale)

~ updated at: 2024-12-27T05:04:37.136Z

More movies like Whiplash, fewer reruns of “Friends.” More math tutoring, fewer sleepovers. More weekend science competitions, fewer Saturday morning cartoons. More books, less TV. More creating, less “chillin.” More extracurriculars, less “hanging out at the mall.”

– Vivek Ramaswamy

Vivek believes these are the times (of acc) where humans need to go harder to push humanity towards Kardeshev Type I and beyond.

~ updated at: 2024-12-27T05:01:21.953Z

I generated this activity aggregate report using https://git-wrapped.com

🏞️ Context #1

~ updated at: 2024-12-27T04:15:03.120Z

I was able to build this feature in my content publisher application quickly (thanks to LLMs)

I can now post to X and Bsky from this app.

🏞️ Context #1

~ updated at: 2024-12-25T12:25:51.557Z

I was able to build this feature in my content publisher application quickly (thanks to LLMs)

I can now post to X and Bsky from this app.

🏞️ Context #1

~ updated at: 2024-12-25T11:34:45.303Z

The urge to rewatch The Big Short for like the gazillionth time.

~ updated at: 2024-12-25T07:44:25.624Z

I found this great explainer by Welch Labs on Mechanistic Interpretability:

https://www.youtube.com/watch?v=UGO_Ehywuxc

I cannot believe I didn’t come across this channel before.

~ updated at: 2024-12-25T07:11:57.775Z

I found this great explainer by Welch Labs on Mechanistic Interpretability:

https://www.youtube.com/watch?v=UGO_Ehywuxc

I cannot believe I didn’t come across this channel before.

~ updated at: 2024-12-24T17:01:19.645Z

Wow! I just came across news that Christopher Nolan’s next movie (2026) is based on Homer’s The Odyssey

Source: https://twitter.com/twitter/status/1871320030773727587/

~ updated at: 2024-12-24T02:43:17.934Z

This doesn’t need to be said but There is so much compressed into this single sentence. Hopes of mankind.

If 2020s had a lesson, I must say this is a good reminder that one can try and learn a lot with A.I. assistants patiently helping you out.

https://twitter.com/twitter/status/1870527558783218106/

~ updated at: 2024-12-22T09:52:44.404Z

Eustress is all you need

~ updated at: 2024-12-22T06:03:16.268Z

Learning in long, uninterrupted flow states.

Executing in short, fast time boxes.

~ updated at: 2024-12-21T13:37:31.068Z

agi has been achieved externally

~ updated at: 2024-12-21T04:20:16.860Z

I’ve always wanted idea threads that easy to use. I try to use Excalidraw to create chain of ideas rendered as bubbles / nodes in a network.

I can’t help but think about the new Pocket hardware that launched this week.

One catch phrase caught my attention. “Conversation threads” if I remember it correctly.

This could help me out but I can’t say for sure.

Sometimes I wanna build it from scratch 🥹 because i genuinely believe being able to navigate idea chains without losing context is crucial in my professional life as of yet.

~ updated at: 2024-12-20T08:16:42.469Z

The cycle of centuries

https://x.com/pertorstensson/status/1869879675507347755

🏞️ Context #1

~ updated at: 2024-12-20T08:05:35.424Z

I use A.I. to warm up the execution process on task chains planned beforehand.

Still majority of time is still spent problem solving (using brain 🧠 ) that finally involving the aforementioned task chains https://twitter.com/twitter/status/1869737000213061648/

~ updated at: 2024-12-20T07:53:48.432Z

My brain still hasn’t fully accepted Nicholas Hoult as Lex Luthor.

Might not be the best casting decision but I hope he proves this wrong.

🏞️ Context #1

~ updated at: 2024-12-20T02:29:53.889Z

I stumbled upon a succinct piece from Paul Irish, about ‘how to enhance website visitor experience on a website using considerable tools at disposal’, while continuing my dive into vercel.

The below points to keep the ‘why’s pickled:

1. Slow TCP ramp up
2. The first 14kb is crucial (because of point 1)
3. Critical CSS for above the fold renders

This talk was referenced in the “7 Principles of Rich Web Applications”.

🔗 I’m attaching the said “Delivering the goods” talk/keynote in this post for reference: https://www.youtube.com/watch?v=R8W_6xWphtw

~ updated at: 2024-12-14T18:00:05.954Z

I found myself revisiting rauchg’s 7 principles of rich web applications:

1. Pre-rendering isn’t optional
2. React immediately to user input
3. React to data changes (server side included)
4. Control data exchange between client and server
5. Don’t break history, enhance it
6. Push code updates (yes client code updates)
7. Predict Behaviour (I honestly think this is overkill — but brilliant if pulled off)

This (re-)read was part of my dive into “how vercel ships good quality software so quickly” assumption.

And oh, here’s the link to the article from 2014, — the examples are really helpful to wrap your mind around the concepts that are seemingly complex (at first).

🔗 https://rauchg.com/2014/7-principles-of-rich-web-applications

~ updated at: 2024-12-14T16:07:15.740Z

Interesting observation:

X has started crawling the “link” in my bio every time i post (or even repost).

~ updated at: 2024-12-14T08:26:46.050Z

Lowkey asking for Tex support on Netnewswire RSS client.

I’ve attached a screenshot of the @NetNewsWire macOS client.

🏞️ Context #1

~ updated at: 2024-12-13T06:30:13.350Z

We can certainly say A.G.I. is on the horizon when Rick and Morty’s Inter-dimensional cable is recreated with A.I. slop.

(It’s almost on the horizon)

🏞️ Context #1

~ updated at: 2024-12-13T05:21:50.452Z

Interesting! but we need to pump up those academic research numbers.

https://x.com/simonw/status/1867362623098171499

🏞️ Context #1

~ updated at: 2024-12-13T03:28:23.713Z

Andrej Karpathy’s book recommends

Science Fiction: All short stories by Ted Chiang Especially: Exhalation, Division By Zero, Understand, The Story of Your Life, Liking What You See, The Lifecycle of Software Objects, What’s Expected of Us The Martian by Andy Weir A Fire Upon the Deep by Vernor Vinge (Chapter one for its portrayal of superintelligence) Flowers for Algernon by Daniel Keyes Atlas Shrugged by Ayn Rand

Nonfiction: The Selfish Gene by Richard Dawkins The Vital Question by Nick Lane How To Live by Derek Sivers 1984 by George Orwell In Defense of Food by Michael Pollan The Accidental Superpower by Peter Zeihan Countdown to Zero Day by Kim Zetter Guns, Germs, and Steel by Jared Diamond An Immense World by Ed Yong The Master Switch by Tim Wu

Fantasy: The Lord of the Rings by J.R.R. Tolkien

Summarized and Compiled by Grok

Karpathy’s Post on X

~ updated at: 2024-12-09T02:36:09.898Z

Stripe Press is a really well designed & implemented website 👏 https://press.stripe.com/

~ updated at: 2024-12-08T10:47:26.468Z

Someone tried to inject a crypto miner into the ultralytics PyPi package (v8.3.41 & v8.3.42)

While these versions no longer are downloadable, if it already is (cached or otherwise) because you’re using it, Please delete those versions / update immediately.

The GitHub issue is still ongoing with what could’ve been done to avoid this. https://github.com/ultralytics/ultralytics/issues/18027

Here’s a comprehensive writeup by the yossarian (@8x5clPW2) on what happened. https://blog.yossarian.net/2024/12/06/zizmor-ultralytics-injection

~ updated at: 2024-12-08T03:06:10.259Z

Sony has expressed interest in purchasing Kadokawa.

https://www.theverge.com/2024/11/20/24300583/sony-kadokawa-acquisition-from-software-anime-manga

~ updated at: 2024-11-24T11:19:19.011Z

Kendrick Lamar’s Link in Bio Page my-gnx.com is hosted on Vercel.

I was wondering if it’s because he was using a link in the bio service. If you inspect the dev console, it appears that there is some relation to Linkfire , but loosely so.

Guillermo Rauch jokes about Vercel’s default favicon still being used

~ updated at: 2024-11-24T02:40:02.070Z

I’ve finally added my 🦋 bluesky profile to my ’link in bio’ page (bagel.ink/@tnvmadhav)

I took this beautiful iPhone Mockup of my link in the bio page using the WebShotPro iPhone application.

Get it here -> bagel.ink/c/DhLwv0

WebShotPro also generates an editable open graph image of the website based on the HTML metatags

🏞️ Context #1 , 🏞️ Context #2 , 🏞️ Context #3

~ updated at: 2024-11-23T12:54:48.910Z

Unread @verge posts are piling up in my RSS client.

🏞️ Context #1

~ updated at: 2024-11-23T04:24:56.840Z

mkbhd is (not lowkey) getting cooked in the comments 🥶🌶️

🏞️ Context #1

~ updated at: 2024-11-22T13:31:46.998Z

Today I learned about the term “Hyrum’s Law” in software engineering.

It refers that even though implementation specifics are hidden, breadcrumbs of implementation will be expected to be consistent when there’s no apparent change the interface signature.

And any implementation change can and will break a small subset of users’ workflow.

And yes there’s an XKCD for it.

Quoting one of Go’s package author:

“I wrote the crypto/rsa comments. We take Hyrum’s Law (and backwards compatibility [1]) extremely seriously in Go. Here are a couple more examples” - via HN

Rabbit Hole [1] https://news.ycombinator.com/item?id=42201892 [2] https://abenezer.org/blog/hyrum-law-in-golang [3] https://www.hyrumslaw.com [4] https://xkcd.com/1172/

🏞️ Context #1

~ updated at: 2024-11-22T06:30:11.909Z

I like the HTTP 418 Teapot joke but I had to do this

🏞️ Context #1

~ updated at: 2024-11-21T12:30:33.678Z

if I already didn’t think that dude was built different, I would’ve been mind-blown 🤯 https://twitter.com/twitter/status/1859378984577876097/

~ updated at: 2024-11-21T09:07:09.668Z

“Mistakes are a portal to creativity” – Cillian Murphy on working with Christopher Nolan and experiences in his film set. He implies that most of brilliant moments on the cinema screen are a result of capturing those lightning in the bottle moments.

• via X

I was fortunate to come across a documentation of one such example in a tweet from a Nolan specific Fan account.

A must follow account for film enthusiasts 👉🏻 https://x.com/NolanAnalyst

~ updated at: 2024-11-18T01:58:58.306Z

John Carmack writes about his (visibly) weird a game engine debugging dream and he suspects that the game engine must’ve been written in C++ and the bug could’ve been a uninitialised variable without doing an RCA.

A level domain knowledge in anything that leaves you guessing causes of uncommon scenarios in most cases is desirable.

• via X

~ updated at: 2024-11-17T06:30:03.034Z

It’s not everyday something like this happens.

Google comeback a certainty?

🏞️ Context #1

~ updated at: 2024-11-15T06:30:09.130Z

Opportunities come and go. You have to catch and squeeze it dry. It’s like catching lightning in a bottle. This happens rarely but you know it when it’s happening. You get into the zone.

Sometimes you feel something strong is approaching but you know you’re not skilled enough to properly make use of the opportunity.

The only solution is to keep improving such that the chances of missing out on sure-shot opportunities due to mere “skill issue” are drastically curtailed.

The next big thing is either approaching quickly or is already here.

my answer to the usual question of “what’s next?”: I don’t know! Because if I did, I’d already be half-way done doing it. And then it wouldn’t really be next, it’d be now.

Finding the next now is the art of wandering, and wandering well takes practice and patience. Don’t rush it.

• via https://world.hey.com/dhh

~ updated at: 2024-11-15T06:30:03.077Z

Opportunities come and go. You have to catch and squeeze it dry. It’s like catching lightning in a bottle. This happens rarely but you know it when it’s happening. You get into the zone.

Sometimes you feel something strong is approaching but you know you’re not skilled enough to properly make use of the opportunity.

The only solution is to keep improving such that the chances of missing out on sure-shot opportunities due to mere “skill issue” are drastically curtailed.

The next big thing is either approaching quickly or is already here.

my answer to the usual question of “what’s next?”: I don’t know! Because if I did, I’d already be half-way done doing it. And then it wouldn’t really be next, it’d be now.

Finding the next now is the art of wandering, and wandering well takes practice and patience. Don’t rush it.

• via https://world.hey.com/dhh

~ updated at: 2024-11-14T18:04:53.148Z

Two huge podcast episodes dropped this week:

— Interview with Dario Amodei by Lex Fridman

It’s about time Anthropic’s C.E.O. got into the limelight to fan the flame that is the narrative of journey to A.G.I. I had recently read his essay and hopefully he talks more about it on lex fridman.

— Interview with Gwern Branwen by Dwarkesh Patel

This one caught me by surprise, I thought he finally did a face reveal when I first saw it in my feed only to know both voice and avatar were modulated.

I’ve been a long time fan of his famous essays at gwern.net and this one is gonna be an all timer.

~ updated at: 2024-11-14T17:07:54.193Z

TIL about a YAML file validation using yq command line tool.

If you have a .{YAML/YML} file, you can run yq eval -v {path/to/file} to do a quick verification on the file format.

https://github.com/mikefarah/yq

~ updated at: 2024-11-13T12:30:03.143Z

I really missed gdb’s coding tweets. I’m glad to know he is back at @openai shipping software.

-X

~ updated at: 2024-11-13T06:23:11.120Z

Damn the latest mkbhd video getting a lot of comments like this but speeding in school zone isn’t just a misdemeanour.

The section of the video was cut out and people are getting angrier because of this.

Ok, he apologised

🏞️ Context #1

~ updated at: 2024-11-13T06:03:28.503Z

“Struggling to document your TILs not because you’re learning so many new things but you’re balancing a backlog load of them and churning mental load in retrospect” isn’t something one should be comfortable with.

~ updated at: 2024-11-06T11:00:02.988Z

notes & excepts from What you know that just ain’t so

We don’t need to know everything for sure before taking action.

[…]Convinced that half the opportunities in the world just don’t apply to your situation and the other half is a slam dunk. Nonsense.

Unlike school and university, where there is reward to those who focus on rigid inputs and outputs, the real world is absurd. The real world doesn’t reward all (apparently equal) attempts equally. There is luck / timing aspect. But the main point remains same, there is trade-off. You MAY become more successful from half knowledge about something than doing it absolutely right.

Don’t take anything as gospel. Learn that every bit of success / failure was influenced by the (seeming) absurdity of the world.

It’s not a bad day (in business or otherwise) if you’re reminded of this.

~ updated at: 2024-11-06T06:30:03.156Z

POV you’re a basic crud app facing a man and his $20/mo cursor subscription.

(he could’ve used a google sheet template but no).

🏞️ Context #1

~ updated at: 2024-11-06T01:50:00.913Z

One can confidently say that they’ve “learned to learn” by breathing the fact that real learning happens through living the reality rather than surfing a commentary on said reality.

~ updated at: 2024-11-05T06:30:03.143Z

I’m experimenting with chatgpt’s ability to count calories of the food and suggest wait time before having the next meal.

1. Click picture of food
2. Send to chatGPT
3. Receive rough time gap in hours
4. Set a timer to wait before having next meal

I’ve initially thought of using it in my local machine but I immediately realized this is useful to me on the go via my mobile browser.

Since I’d already had a web tools application provisioned up and running. I’d used chatGPT to convert the flask app to a @golang HTTP handler.

I’ve now exposed it via the web .

Bring your own chatGPT API Key!

🏞️ Context #1

~ updated at: 2024-11-05T02:48:45.828Z

One can confidently say that they’ve “learned to learn” by breathing the fact that real learning happens through living the reality rather than surfing a commentary on said reality.

~ updated at: 2024-11-04T14:49:50.662Z

If this doesn’t make you go “hell yeah” or “hell no”, you’re not on the spectrum because there is no spectrum here.

via X

🏞️ Context #1

~ updated at: 2024-11-04T14:30:17.297Z

The funny thing about the human perception of hard work is that the apparent hurdle to subsequent attempts to utter failures is not a psychological one.

~ updated at: 2024-11-04T13:00:02.953Z

Can you solve your way through a problem through sheer spaced repetition?

~ updated at: 2024-11-04T06:30:03.510Z

A developer knows how straining and difficult it is to document a debugging session especially in hindsight where a certain optimum spurs out of the void that all suboptimal attempts look pale in comparison so much so that sometimes you start feeling a little embarrassed (I do).

I enjoyed reading and appreciate this post on improving memory usage performance by @b0rk. Also learned about many packages along the way.

I’d like to practice doing this more often.

~ updated at: 2024-11-04T06:11:31.292Z

The funny thing about the human perception of hard work is that the apparent hurdle to subsequent attempts to utter failures is not a psychological one.

~ updated at: 2024-11-03T13:05:54.130Z

Can you solve your way through a problem through sheer spaced repetition?

~ updated at: 2024-11-03T12:43:53.373Z

Clarity breeding action doesn’t take into consideration the baseline habitual tendencies.

~ updated at: 2024-11-03T12:30:05.713Z

Fleeting thoughts influenced by the spark of moments in the universe instead of steady build up of logical implications is the flywheel that trades your time for life experiences.

~ updated at: 2024-11-03T12:30:02.989Z

A chunk of my public content feed via my content publisher system goes along with media (mostly images). Sometimes the media attachments contain a significant amount of context that may be lost with text only compilation.

So I decided to send static media links of images/assets hosted on my content publisher application to my public feed .

This feed is used by several feed renders that are live on my blog site [1] , my GitHub profile [2] and my post archiver cronjob.

🏞️ Context #1 , 🏞️ Context #2

~ updated at: 2024-11-03T03:30:23.369Z

A small % of iPhone 14 plus phones have a camera hardware issue.

Apple says it or an Authorized Service Provider could service your phone for free.

Some conditions apply but you can check with your iPhone’s serial number on Apple’s support page

via The Verge

I’ve had a similar issue but with the iPhone 13 series where the display hardware was faulty and Apple Authorized Service Provider verified and fixed it for free.

I’d written and shared a post about it and you can read about my experience on my blog .

~ updated at: 2024-11-02T08:30:05.489Z

A successful founder gets things done.

A person who has an intense want will find ways to get things done.

This intensity comes from life experiences.

Bad childhood can build character and lead to strong wants.

See where the road converges? https://twitter.com/twitter/status/1852017824971264177/

~ updated at: 2024-11-02T01:39:09.985Z

I spent some time to make some significant performance improvements (low hanging) to my blog.

The lighthouse score on the mobile looks pretty good now.

I’d also run a scan on @cloudflare’s radar tool and found that total bytes score was less than 512kB.

This means that my little site has (unofficially) entered the 512kb.club 🥳

I opened a pull request to make it official. Fingers crossed 🤞🏻

~ updated at: 2024-11-01T15:46:09.891Z

I’m experimenting with chatgpt’s ability to count calories of the food and suggest wait time before having the next meal.

1. Click picture of food
2. Send to chatGPT
3. Receive rough time gap in hours
4. Set a timer to wait before having next meal

~ updated at: 2024-11-01T12:29:27.349Z

what? nothing. just me thinking about “the boy and the heron” / “how to live” an entire week after watching it. i thought i knew hayao miyazaki but he surprised me with this incredible feat. again.

~ updated at: 2024-11-01T04:41:08.425Z

TIL we can prevent bash script from prematurely exiting using set +e.

set +e
<some command>
set -e

This knowledge is useful if you want to run certain scripts and perform custom actions on the resultant exit code without losing information due to premature exits.

From bash reference manual

~ updated at: 2024-10-30T12:43:30.749Z

[1] Open Source AI Definition as per Open Source Initiative

The OSAID offers a standard by which community-led, open and public evaluations will be conducted to validate whether or not an AI system can be deemed Open Source AI.

“The new definition requires Open Source models to provide enough information about their training data so that a ‘skilled person can recreate a substantially equivalent system using the same or similar data,’ which goes further than what many proprietary or ostensibly Open Source models do today,” said Ayah Bdeir, who leads AI strategy at Mozilla”

[2] Open Source AI Definition 1.0

1. Free to use without permission
2. Study the components to learn how it works
3. Modify for use to change the output
4. Share with other with or without modification

These freedoms apply both to a fully functional system and to discrete elements of a system. A precondition to exercising these freedoms is to have access to the preferred form to make modifications to the system.

This involves data, code and weights.

~ updated at: 2024-10-29T06:30:05.879Z

This is a dumbed down (simple) take.

A.I. is to “building” as what client side javascript is to HTML.

Elimination of extra time and effort to gain a certain outcome. https://twitter.com/twitter/status/1850940029155373382/

~ updated at: 2024-10-29T06:30:03.058Z

I was able to convert, and load some of my computer science related notes in .txt files like HTTP and Redis onto Anki in seconds.

I was able to revise in matter of minutes!

Thank you @openai :)

One can find this script useful if you wanna automate Anki flashcard generation with Python and ChatGPT. https://twitter.com/twitter/status/1850163990296433051/

~ updated at: 2024-10-27T06:30:17.002Z

TIL about the cookiecutter PyPi package

It’s a quick way to scaffold new python packages, plugins, django projects and many more from pre-built templates.

Using jinja2 template syntax, one can create an entire cookiecutter project template for use and reuse.

A comprehensive documentation can be found on the internet .

~ updated at: 2024-10-26T13:30:03.046Z

AI is (unsurprisingly) good at converting technical notes into Anki Flashcards.

For the uninitiated, it would’ve taken several minutes to construct questions and answers after taking hours worth of notes but with a relatively small LLM like GPT-4o-mini, one could churn out Q&A list in seconds.

I believe certain unnecessary hurdles of the theoretical learning process is eliminated.

And yes, all one needs now is to be able to actually import said compilation onto Anki in bulk. And most importantly, eat the damn frog!

It seems like AnkiWeb doesn’t support export or imports.

I’ve installed Anki Desktop app so I could bulk import new flashcards.

I’ve generated a simple python script to process a text file using GPT-4 (customize prompts) and output a anki importable .txt file.

The process of taking a note file and generating flashcards is now streamlined.

I could setup a local cron job to run everyday, that processes the notes from my local notes directory for the day and generate anki flashcards with the learnings.

I could configure a LaunchAgent or Daemon perhaps using .plist configuration.

~ updated at: 2024-10-26T13:14:16.350Z

If you don’t know how something impacts the world around you, you don’t fully understand. https://twitter.com/twitter/status/1849773180136599989/

~ updated at: 2024-10-26T06:30:03.189Z

TIL that ICE TSUNAMIS are a thing!

https://en.wikipedia.org/wiki/Ice_shove https://twitter.com/twitter/status/1849163888895599031/

~ updated at: 2024-10-24T12:07:14.479Z

To the builders of the future, you need to relive the pain that the thing has genuinely relieved.

~ updated at: 2024-10-24T06:30:02.956Z

Excerpts from “Desperation Induced Focus”

“Most big companies aren’t focused on creating things out of nothing. Someone else made the magic money-making machine, and they assume that it will just keep working.”

“This lack of focus is a luxury and a disease.”

This is called peace time thinking. Thoughts generated out of seemingly derived notions or non-axioms. Sometimes too much trust in the system.

My advice to people when they are thinking about instituting a new process is to go to a whiteboard and write down the answer to this question: “If you could only get one thing done this year, what would it be?”.

Desperation inspires creativity and intense focus. It is an essential ingredient to building great products and services.

So, the next time you feel desperate, lean in. Embrace it. Use it as the fuel to create the next founding moment4 for your company.

And the next time someone tries to tell you to do something because a big company does it, be suspicious.

One must arrive at something from first principles and not because someone told them (a.k.a. by proxy). Keep the advice in mind and use it as a reminder but never an axiom.

To the builders of the future, you need to relive the pain that it has genuinely relieved.

~ updated at: 2024-10-23T17:30:03.050Z

I’d stumbled upon a rather brilliant @leetcode editorial piece and a useful mind map on solving problems.

This something that a younger version of me would’ve appreciated especially when starting out.

I’ve shared some notes and excerpts on my blog

~ updated at: 2024-10-22T03:37:31.352Z

Ownership is important. Start your own blog! https://twitter.com/twitter/status/1848261668024807887/

~ updated at: 2024-10-21T12:30:03.083Z

Notes from leetcode’s editorial section on solving data structures and algorithm problems in general.

After solving a leetcode problem, found myself checking out an editorial section to see what other approaches I could’ve missed. What I found was unexpected.

A guide to identifying patterns and choosing approaches. This is something I could’ve used when I was a couple of years younger just starting out.

So this time before diving into the answer, let’s understand a few general patterns that you can use in your future journey:

Sorted Input:

Apply binary search for efficient element lookup. Use the two-pointer technique for problems involving pairs or segments. Unsorted Input:

Apply dynamic programming for questions related to counting ways or optimizing values. Use backtracking for problems that ask for all possibilities or combinations (this is also a suitable fallback if dynamic programming isn’t going to work). Use a Trie for prefix matching and string-building scenarios. Use a hash map or set to find specific elements quickly. Implement a monotonic stack or sliding window technique for managing elements while continuously finding maximum or minimum values. Input is a Graph or Tree:

Use DFS to explore all paths or when the question does not require finding the shortest path. Use BFS when the question asks for the shortest path or fewest steps. For binary trees, use DFS if the problem involves exploring specific depths or levels. Linked List Input:

Use techniques involving slow and fast pointers or “prev” and “dummy” pointers to facilitate certain operations if you are unsure how to achieve a specific outcome.

Note: There’s so much more to this pattern! We just wanted to give you a glimpse of what pattern recognition boils down to in its simplest form. Feel free to add your own flair and create a detailed chart!

Here’s a useful mindmap of techniques to use per problem type

~ updated at: 2024-10-21T03:50:20.239Z

This takes the general concept of “put your money where your mouth is” and zooms out a bit IMO.

Reframe as “if your mouth isn’t where your money is, where is it?” https://twitter.com/twitter/status/1848025357061926957/

~ updated at: 2024-10-20T22:00:02.964Z

Perhaps goals are just misaligned.

It’s not about the immediate output but how it interacts with the world.

Zooming out every now and then, helps in keeping tunnel vision at bay.

~ updated at: 2024-10-20T18:30:03.007Z

Moving slowly is much better than giving up.

~ updated at: 2024-10-20T15:44:28.929Z

Notes from Local First Software :

With the introduction and proliferation of “cloud-hosted” Software as a Service (SaaS), especially for consumer software, it becomes unclear who owns the data that flows through them.

Many times, the users who upload/generate data don’t have complete control over what happens to them.

Moreover, the experience of using application software has become unnecessarily complex and painful.

If you’re a software engineer, designer, product manager, or independent app developer working on production-ready software today, how can you help?

The paper proposes an alternative concept called “Local First”, a guide for software developers to support experiences w.r.t. speed, ownership of data, and quality of use.

Seven points to use as guideways for the uninitiated:

1. Make it Fast

Feedback has to be instant and non-blocking. Network calls (to servers or stores for example) affect this the most.

2. Multi-Device

As the name suggests, “local first” doesn’t mean data and software should always be offline; rather, the “sync” if necessary, needs to be quiet, in the background, and non-blocking.

3. Offline Support

Not all software / client-side applications, need to interact with remote servers for all tasks.

Offline support means having a copy of all relevant resources available to work on without internet access.

4. Collaborative

When a large enough ecosystem supports local first development, there needs to be a system to support collision detection and resolution systems.

This is highly relevant for Google, GitHub, etc.

5. Longevity

This is purely data-centric. Data needs to persist independently of the software that creates/processes it.

This means data should be importable/exportable into standard formats.

6. Privacy

Data that is private must be able to remain private. Local First Development supports this philosophy.

7. User Control

The user must be able to choose where the data is stored. The user should be able to have 100% ownership of the data. The user can choose to do whatever they want with it.

~ updated at: 2024-10-19T08:34:42.678Z

Great timing.

I looked forward to the customizability of NotebookLM’s, audio podcast generation (Solid Product!).

I can now customize and provide instructions on what to focus on when given a particularly technical document.

Thank you @Google and team! https://twitter.com/twitter/status/1846946225251406039/

I can’t wait to start implementing this today!

~ updated at: 2024-10-18T03:38:04.022Z

TIL you can inspect a CLI command using the type <cmd> CLI command.

I usually use which <cmd> to verify this but type provides a human-readable description/comment of the command.

Here’s an example from my terminal:

$ type openai-env
openai-env is a shell function from /Users/<user>/.zshrc

$ which openai-env
openai-env () {
	export $(cat $PYTHON_PROJECT_PATH/.env)
}

I learned this from Julia Evans’ recent comic strip on PATH

~ updated at: 2024-10-16T03:28:30.965Z

Excerpts from Machines of Loving Grace by the CEO of Anthropic

…“I think their rate of discovery could be increased by 10x or more if there were a lot more talented, creative researchers. Or, put another way, I think the returns to intelligence are high for these discoveries, and that everything else in biology and medicine mostly follows from them.”

Yes, it’s mostly not (just) about doing what we already do… just faster. But, having enough brain juice (left) to be open to doing other things.

“While that might sound crazy, the fact is that civilization has successfully navigated major economic shifts in the past: from hunter-gathering to farming, farming to feudalism, and feudalism to industrialism. I suspect that some new and stranger thing will be needed, and that it’s something no one today has done a good job of envisioning.”

I try to believe that cost of executing and applying “solved problems” will go down eventually with machines taking over the replication part.

It’s then up to humans to pick up new challenges. This could be improving lifespan, brain capacity, low latency and efficient modes of communication, transport, energy efficiency, climate control/repair, food and water scarcity, etc.

To know about what lies beyond the horizon (which appears to be approaching quite rapidly), we must practice research ourselves, observe the work of researchers right now. What they think / dream about.

Human creativity will have to be put to the test to find meaning on this pale blue dot .

~ updated at: 2024-10-13T12:30:03.176Z

I’m (obviously) not there yet but doing stuff like this makes you competent.

You’re either doing it for fun/satisfaction/requirement (or) you’re doing it to learn.

Either way you’ll gain enough skills to pick more ambitious projects.

https://twitter.com/twitter/status/1845186558414159875/

~ updated at: 2024-10-13T09:13:33.290Z

I’m reminded of Kernighan’s Law .

Debugging is twice as hard as writing the code in the first place. Therefore, if you write the code as cleverly as possible, you are, by definition, not smart enough to debug it. ~ Brian Kernighan

But here’s the thing, if you stick to simple solutions, you’ll not be prepared for inevitable scenarios where your debugging skills aren’t up to the mark.

There is great risk in coming up with clever solutions but this actually pays off when you constantly push this boundary by debugging and finding faults in said clever solutions. https://twitter.com/twitter/status/1845062132326449455/

~ updated at: 2024-10-12T12:30:03.243Z

Linus Torvalds on working with GitHub Pull Requests

~ updated at: 2024-10-04T06:30:03.070Z

TIL that the .me is country code based TLD (ccTLD) for Montenegro.

And that it’s not some individualistic / personal branding thing that we could take as granted.

Reference: Wix Perplexity

~ updated at: 2024-10-04T04:30:09.183Z

I finally bit the bullet and learned to use SwiftData and it turned out to be simple to use (as a beginner).

I’m playing around to try and learn how the data is modeled.

I could go from ephemeral to persistent storage in a couple of hours.

~ updated at: 2024-09-29T14:59:30.158Z

Excerpts from Using Progressive Enhancement by GOV dot UK

“Progressive enhancement is a way of building websites and applications based on the idea that you should make your page work with HTML first.”

HTML and CSS layers are fault-tolerant (browsers will ignore declarations that it doesn’t understand).

Javascript must only be used to enhance user experience in necessary parts of data that are already in HTML.

“Where possible the JavaScript should enhance HTML and CSS that provide the same core functionality. For example, an autocomplete could enhance an element, or something similar. This still lets the user do what they need to do, even if the JavaScript fails.”

HTML and CSS offer the bottom layer (Layer 1) where things can happen but rather slowly. Javascript layer to exist to make a certain action faster. If the javascript layer is to be removed, the required functionality would still be possible albeit slowly as initially intended.

This way, no process is “blocked” during Javascript-based failures, etc.

“If you believe your service can only be built using JavaScript, you should think about using simpler solutions that are built using HTML and CSS and will meet user needs.” For example, if you want to use JavaScript to provide interactive graphs, other options are to:

Display the data in a table Allow the data to be exported so that it can analysed in another application Pre-render the graphs as images

If the core functionality of your service cannot be provided without JavaScript, you’ll need to consider how users can access your service through other channels. This might be telephone calls or in-person visits to offices.

Looks like the UK gov prefers in-person processes rather than injecting Javascript into government websites. This a highly opinionated decision.

If you do choose to use client-side JavaScript frameworks, be aware that although they can be helpful when building a service with a complex user interface, they can introduce problems.

Using a client-side JavaScript framework can:

Increase the overall size of your code base and push processing to the client side, causing performance issues for users with a slower network connection or lower-powered device Create a reliance on third-party code that your developers do not have control over, requiring you to make major changes to your service to stay up to date with changes in the framework make it difficult to find people with the skills required to maintain the code, if the frameworks lose popularity over time If you use a JavaScript framework you should:

Be able to justify with evidence, how using JavaScript would benefit users Be aware of any negative impacts and be able to mitigate them Consider whether the benefits of using it outweigh the potential problems Only use the framework for parts of the user interface that cannot be built using HTML and CSS alone design each part of the user interface as a separate component

Having separate components means that if the JavaScript fails to load, it will only be that single component that fails. The rest of the page will load as normal.

I can understand this approach, keeping things light and simple is a win, and keeping the user base in mind (literally everyone in the country in this example), it’s important to ensure stability and dependency-less design.

The userbase for a normal corporate may or may not be the same, so the design for the userbase is a good way to think about problem-solving.

“If you use JavaScript, it should only be used to enhance the HTML and CSS so users can still use the service if the JavaScript fails.”

“Some users may deliberately turn off features in their browsers. You should respect their decision and make sure those users can still use your service. Some users may deliberately turn off features in their browsers. You should respect their decision and make sure those users can still use your service.”

A user can turn off JavaScript in their browsers AND direly need certain services.

“Do not build your service as a single-page application (SPA). This is where the loading of pages within your service is handled by JavaScript, rather than the browser.”

A subset of users using accessibility features may get a sub-optimal or dead-end experience. The heavy use of back and forward buttons and page refresh can cause states that they don’t intend or understand.

The UK government has written another piece on why progressive enhancement is good (I agree in favor of incremental improvements as good engineering practice):

Quoting architect of gov dot uk

“Progressive enhancement is about resilience as much as it is about inclusiveness.”

~ updated at: 2024-09-29T04:39:18.497Z

TIL there is a site pyvideo.org that curates videos from community and related events.

~ updated at: 2024-09-28T16:30:03.005Z

I found a really useful refresher/guide to get back to while working with SAML

Worthy of a bookmark!

~ updated at: 2024-09-28T12:30:03.033Z

Some text snippets from the white paper on reading comprehension on plain paper vs on a screen

“If texts are longer than a page, scrolling and the lack of spatiotemporal markers of the digital texts to aid memory and reading comprehension might impede reading performance.”

“…Hence, we cannot say whether the visual ergonomics of the laptop computer screens hurt subjects’ reading. In future studies, employing online measures including eye tracking methods would appropriately address this important issue.”

“…. Scrolling is known to hamper the process of reading, by imposing a spatial instability which may negatively affect the reader’s mental representation of the text and, by implication, comprehension.”

via HN

~ updated at: 2024-09-28T09:30:03.016Z

If your AWS deployments seem unnecessarily slow, you could do something about it.

I’ve found a well-written piece that could help.

~ updated at: 2024-09-27T12:30:03.052Z

They’re perfect for airpods https://twitter.com/twitter/status/1839351084630028749/

~ updated at: 2024-09-27T06:30:03.003Z