# Rahul Tarak
> Builder, founder, and engineer. Loves solving deep technical problems.
## Writing
### really building for an AGI future?
*November 14, 2024*
**Tags**: startups, ai
> framework for thinking about agi "proof" startups
import { NoteReference } from "~/components/NoteReference";
sorta rambly thoughts but "framework" for thinking about building startups ready for a [powerful ai](https://arc.net/l/quote/xebyehid); most companies ik and founders i talk to seem to just skirt this problem, and i feel like that was fine to do last year more so but increasingly less fine
i've been finding it hard to see through the fog of war what the future looks like in terms of what are real durable business and not - this is an evolving framework to think about how to build for that
ideas/frameworks to borrow from
- bezos has this quote to the effect of everyone asks what do you think will be different in 5/10 years but the question to ask is what stays the same - this is what you can build a business around
- i love both frameworks of "[marginal returns on intelligence](https://arc.net/l/quote/usalpydw)" and "[compressed 21st century](https://arc.net/l/quote/bkqmrwdy)"
in software, outside the labs or building models to compete where the labs aren't focused, there is a knife edge to walk on where you don't get eaten by the model provider and don't get so easy to build with the existing model that the enterprise does it themselves (this is before use the model for codegen) - i find most ai agent type companies to fall of this knife's edge.
two notes
1. i'm thinking of baseline software commoditizing as a reaper that increases competition, lower prices but not necessarily death of business
2. the knife edge is thinking about how to not be collateral damage from models - i think here it is better to compete head-on and navigate an attack vector that not thinking clearly about it and being collateral damage
even my fav ai company (non lab) - cursor with insane pmf is not obvious to me that have a durable 5-10 year business; _at some point codegen is so good, the comparative advantage in writing code is gone (can be system architecture or product or design but not writing code), before this the codegen is good enough it is like reviewing a pr - ideally on the web._
karan raises a counterpoint that models tend to get to 80/90% on harder and harder tasks before saturating 95-100% on even simple task - this would give the comparative advantage back and need for an ide
another point is cursor could evolve into that next ux on the web or something better.
i think there are some companies walking this line and sure some will make it out but feels like a very tight line to walk. maybe [composio](https://composio.dev/), [browserbase](https://www.browserbase.com/), … ?
_tangent ramble:_ it doesn't as obviously feel like some traditional software companies are on such a collision course with AI.
some obviously are, just through more competition less prices - this is mostly through commoditising the ability to write that software. that being said this doesn't feel as **acute** as being on a collision course with the lab directly or the enterprise integrating the labs work.
examples here to explain what i mean - it doesn't look like [bun](https://bun.sh/) or [astral](https://astral.sh/) business is meaningfully affected (to whatever extent they had a business) but what about vercel? i could see them being fine? but also maybe dx doesn't matter in this world so they aren't? feels like stripe is fine.
expanding on what stays the same, some ideas, some concepts, some businesses
- airbnb most probably - we still probably want vacations vs uber that seems pretty dead with waymo and tesla eating their lunch, probably still been fine for a while but will whimper and taper out
- network effect business here are interesting as probably have low marginal returns to intelligence in breaking them
- stripe most probably - unless no income weird ubi, probably paying for things, maybe bridge or whatever is the agents way to pay for things
- sports - nike wouldn't really be in trouble, have had stockfish for years don't really like watching stockfish play chess
- music - think it is super important to be a shared meme, i think ai music tooling to making shared memes is interesting but not sold on the personalised music(vs very sold on personalised software)
- movies/tv - blockbusters probably the same, shared meme; netflix slop can probably be ai generated and personalised
- games - closer to movies some shared meme, some personalised.
- human connection stuff - wanting friends, love
- probably decent durable business here
- … ? _should expand this list much longer_
- maybe example from conversation is underwriting insurance for AI Agents
### _tangent:_ opportunity cost in a revolution
there will be many things take make $1M, $1M/year even $10M/year but in the last wave of this scale - there were the real trillion dollar market opportunities or at the very least many 100+ billion market opportunities. i think the last 10 ish years didn't have this scale of opportunities outside deep-tech go build the ai revolution
this is all to say, choosing market in a revolution is super important cause the size of the pie up for grabs in enormous - this is much where you run into the steamroller
note: i hate the stay in your lane sorta advice, find some tiny niche or crevice where the steamroller doesn't touch and beg of scraps. imo startups are about building generational iconoclast companies that disrupt industries not begging for scraps in front of a steamroller.
if you are thinking in these lines and want to chat or collaborate, find some time to [chat](https://cal.com/cryogenicplanet)
---
### Scalar retro and post-mortem
*March 1, 2024*
**Tags**: startups
> a post-mortem of scalar
import { FloatingNote } from "~/components/FloatingNote";
import { NoteReference } from "~/components/NoteReference";
import { ToolTipNote } from "~/components/ToolTipNote";
import { ClientTweetCard } from "~/components/ui/client-tweet";
_i've rewritten this post some many times, but most of it was in feb 2024 but updated in dec 2024_
[scalar](https://scalar.video/login) was a professional collaborative video editor think of it as a figma for video editing.
this is going to be long, and a lot of story telling.
## The leap
I spent the last three years trying to beat adobe at their own game, why would I do something so stupid? I'll explain but first a bit of context.
This story starts in 2020 at University of Toronto, and I really hated it there. But when the world shut down in March 2020, it was the perfect excuse to convince family to let me take a gap semester. This was largely due to a close friend Arnav. He had taken a gap year before uni and had showed me how it could be done.
This was the first large bet I had taken on myself, I ended up not picking any classes for the fall and knew I had spent my time well so I immediately installed a linux distro where I couldn't play video games. This is when I started experimenting with `WASM` which was really new. Many linux distros have license problems with video codecs especially `h.264` and I wanted to make a quick UI for `ffmpeg` to solve video workflow tasks I did; without having to pull my hair out. This lead to building [videotranscode.space](https://news.ycombinator.com/item?id=24588523), it was an open-source attempt to get `ffmpeg` to work on the browser.
One morning it blew up on [Hackernews](https://news.ycombinator.com/item?id=24588523). Suddenly we had 15k users. During this time, I was participating in [Pioneer](https://pioneer.app/), at the time an infinite tournament. It was awesome I loved posting updates every monday and getting feedback from the community. The Pioneer leadboard really helped me hone my product sense.
To my absolute surprise, I got accepted into [Pioneer in October 2020](https://pioneer.app/blog/meet-the-pioneers-october-2020/). At this time I was still a college kid building an open-source project looking for an internship. But here Pioneer changed the trajectory of my life, those first couple of office hours with Daniel and the team were incredibly motivating and lit an ambition in me.
While I am storytelling here, I immediately want to call out to big takeways
- How valuable it is to see people around you do ambition/interesting things,
I would have never even started without seeing{" "}
Arnav
. When you see Zuck he is an infinite distance away; for me I knew Arnav was
better engineer and smarter than me but 15/20%. I can overcome 15/20%
- How life changing successful people just believing in you is; Daniel didn't spend that much time on me especially then maybe thirty dedicated mins and a hour or two in office hours but just that belief and backing triggering the internal ambition in me.
## Zack
[Zack](https://twitter.com/zack_overflow) is a vital part of this story. I met him in college where we hacked on random shit together in our school's startup incubator. Even at this time Zack was an incredible engineer, over the course of the next couple of years morphed into a true 10x engineer. I've always said Zack was our Evan Wallace, and incredibly critical for the kind of product we were building.
When Zack mentioned he was taking a semester off in early 2021, I knew I wanted to work with him.
During this time period, we were purely motivated by curiosity and technology; just wanted to push the limits of what we could do with WASM and video. During this period we were working on a client side video editing sdk: [api.modfy.video/](https://api.modfy.video/)
We really wanted to make a video editor, but weren't particularly confident in ourselves yet. So we decided the next best thing would be [client side components for editing](https://docs.modfy.video/docs/components/list/trim/). The sense was customers faster building out this SDK while still building the same core tech needed for the video editor.
## The video editor
The core inspiration for a collaborative video editor came from my experience as an amateur film maker. I was making a 20min long short-film after high school and I used to drive an hour and half everyday to be able to edit with my team (my friends). This had always been my experience with editing short films we made. **We always had to edit in-person.**
The core thesis came from my personal experience but we validated this problem by talking to hundreds of video editors. They ranged from small to medium creators, freelance editors and later even some folks in hollywood.
**The vision for the product was to build a real professional class collaborative video editor. A product with a easy to make something decent}>low floor, a high ceiling - one you can use 8hrs a day. A figma like experience, one where it is seamless to have sync or this would be like send a figma link and get an immediate dm on slack or then they jump on a call}>soft async experiences something that is really not possible in video editing workflows without being in person.**
We thought this was finally possible because of `WASM` which would let us do everything client side, providing an excellent performant experience like figma had. We were especially fascinated by using `WASM` to render videos locally. Our thinking was the unit-economics would be much better and would provide an incredible free-tier experience. This was a mistake for a few reasons; First trying to render video locally was just not performant or stable enough. Second, users didn't really care about this and we were optimizing for unit-economics too early. Finally we were just nerd snipped by `WASM`. This mistake cost us about a year of our lives.
We wanted to rethink how video editors could look if they were built from the ground up for collaboration. A lot of ideations of user interfaces and experiences came up during our [Moonshot sessions](https://arc.net/l/quote/qmnailaw).
There were many many experiments wanted to share a few highlights here, these sketches com from a 10min sketch idea a friend gave us for expressing some of these moonshot ideas. All of the amazing design work is curiosity of [Malay](https://twitter.com/malayvasa/), our excellent designer.
There are countless iterations that I could go on for days about. I find the UX iterations a bit easier to consume but there is plenty on the tech side too. Maybe Zack or I could write an entire other post about that side.
## Trying to hack GTM
from future rahul here: will get into this more in reflections but this was
a bad idea to even try to do
We knew the product we wanted to build would takes years, and we really wanted to find ways to hack GTM and be engaging with our users well before our product was ready. Not sure if this pursuit itself was a good idea or not but nevertheless we tried many times.
The first of these was Anno.so, this was early 2022 and we had just started On Deck Accelerator - similar to YC}>ODX and wanted to see if we could get users who were already collaborating using us while we worked on the core editor. Tbh we did a great job with anno initially, it was built and launched completely in two weeks - yes it was just a basic frame.io clone but simpler but we good good traction initially with it. We had film students using us, some creators and even netflix india for a minute - but most of these users never wanted to talk to us and churned 2/3 months in. Moreover while we though, sharing a lot of code with anno would make sure it wasn't a distraction it definitely split our focus and ended up making poor progress on both ends.
As more physical events started happening in 2022 again, we realised a good place to get user feedback and talk to users were film festivals. The problem was our video editor was nearly stable enough → remember our product didn't work for a user if they could use us for 15/30 mins a day, it was their daily driver. So we weren't selling a just a leaky bucket but a bucket with no floor. While this festival was a great motivation for the team, as we got such a positive response from 100s of editors and filmmakers in person, overall it wasn't a very fruitful endeavour.
{/* < embed tweet here https://x.com/zack_overflow/status/1586288992680493057> */}
## Not finding PMF
While we started in 2021, 2022 was our like first proper year with real capital and alignment on what we were building, by the end of it we felt all over the place and nowhere near reaching PMF. Zack and I sat down for hours and wrote a detailed reflection of that year, and the plan for the next. Thankfully our visa stuff had been sorted out, so we both flew to SF with a much clearer goal.
Around this time, we made some keep changes to our strategy too.
- we dropped `wasm` entirely and decided to drop the focus on a good free tier experience
- we wanted to introduce a new video editing experience which we called "infinite canvas", we thought this was better for collaboration but also by this can be a whole blog post, but it is a super interesting thing about product experience is if you put a timeline in front of a user they expect everything they want from a timeline to work but if you change what the base experience is they are willing to play more}>moving away from an expected user experience users gave us more leeway.
- we also did a full rewrite of the product and it actually made us move a lot faster, it cleared out all the technical debt.
the default advice you here is move fast mvp everything but for products like this, it creates lot of technical debt and sometime spending the time to rewrite it to get the abstractions write is good. (more on this in reflections)
These were some of our best months, we were shipping with a small group of users in a filmmaking discord, talking to 100s of users a month, having email threads back and forth with users.
## It all falls apart
While things seemed to be going well at first, we knew we were in a sticky position - we had to build for another year or two to get to product to a truly usable state for **8 hrs a day** and the video editing space had become hyper competitive.
Around this time is when GPT-4 started this new AI wave and it was clear to us that we had to stay on top of AI enabled editing features as well, we felt that if we had a good base editor we could easily go after this but fighting on both axis seemed like a losing battle.
So we decided that we had to close a much larger seed round or pivot - we pushed for this round but went into the classic fundraising death spiral. as we focused on fundraising, the product lost momentum and vice versa. In all honesty, retrospectively we probably should've pushed harder here but we got jumpy.
We had done a few small/soft pivots inside video by this point, but while considering hard pivots it was clear Zack and I were interested in different things and decided to just work on what we found interesting.
A fair question to ask here, is why didn't we sell? get an acquisition or an acqui-hire? We had a few offers as early as 2021, and even had a serious offer around the same time in 2023. Tho this offer was when things were going well, but regardless we both sorta felt like doing our own thing and not necessarily joining another org.
The acqui-hire would've only really helped in saying "oh i sold my last company".
## Feelings of failure
this gets better overtime, but also in conversations in the last year with
so many founders i realized this is a common feeling
_i promise i will get to the reflections and lessons soon_
I loved working on Scalar, it had some of the best moments for me and I really wanted to solve the problem of having this seamless collaboration. It sorta broke me to see myself fail, i knew the statistics but it always felt like we/i were different - that we would win at any cost. Matt Dameon's character - "The truth is I never really considered the possibility that my planet wasn't the one" - it my head this read more like "I never really considered the possibility I wasn't the one" (here the one meaning the one who wins)
I loved my team - Zack was a truly incredible engineer and great thinking partner. Malay was an incredible designer and thought me so much, we had these incredibly deep conversations about design and product. Geek was a great engineer and a blast to work with. And to so many others that contributed over the years.
It was also sort of my identity and if i am being honest sort of a proud status symbol i wore - so i was lost what was i, if not a founder?
this very specific to me and how i think about being valuable in the world
So I've talked a bunch about how for me the only point of anything to create value at scale - if I am unable to do so what should I do with my life
There are two branching possibilities, actually three -
{" "}
- Keep trying to create value at scale - challenging the core conceit and
probably the most heavily but difficult framing
- Give up on your principles and just be happy while being unable to create
value at scale - just be kumbaya, idk if this is me or even something I want
- Just get a regular boring job and be miserable while retaining your
principles - which probably slowly fade over time
## Reflections
I've spent so much time thinking about this and made so many arguments to people as i explain what went wrong to them, here is what i think are the real things that we did badly.
- We were wishy washy about the type of product we were building, we knew what we wanted to build and had conviction about the problem but not on our approach to company building. We didn't really call our shot.
This lead to poor decision making overall, we let the noise get the better of us - listened to advice for different type of companies and products. But primarily it let us make poor product and engineering decisions that were not suited for the type of company we were building.
The key example here, there is no world in which we should've MVP'd our multiplayer engine but we did cause we were in the MVP and move fast for everything stage - this cost us an year and a half.
- To extend on that point, for "deep software" product you have to nail technical abstractions that compound the way you would for deep tech companies. This is not something to MVP at all.
- Bad relationship with capital, we didn't think we needed it as much as we actually did. And we didn't use the money we had well, not aggressive with it and very skiddish - we were too cheap.
- Remote sucks! it works for some people but it really sucked for us - we had to be in person for this product to work (and ideally in sf were late to this train too).
- More a finer point, but i think we should've spent more time prioritizing the co-founder relationship specifically when things were bad - we were both young, immature and definitely harder on each other than we should have been.
## End credits
_i had a what's next section but so much has changed since i first wrote this that it doesn't make sense anymore, just follow [Zack](x.com/zack_overflow) or [my twitter](x.com/cryogenicplanet) for the latest_
There are an incredible amount of people that have supported me across this journey, some have been mentioned here but there are countless many more - this was one of my favourite aspect of the silicon valley culture, just the amount of people that were incredibly helpful was so immense.
I try as it is to pay it forward as much as I can, and would love to help in any way I can. Feel free to reach out, have an open cal link [here](https://cal.com/cryogenicplanet/) and my email is `hey@cryo.wtf`
---
### Goal: Increase Energy Consumption
*September 9, 2024*
**Tags**: philosophy, economics
> A "good" goal for humanity to increase energy consumption
import { NoteReference } from "~/components/NoteReference";
_**i**'ve made this argument before but want to write it down in one place, none of this is that ground breaking or deeply original - just when i talk to more "normie" friends they don't fully get it and this is trying to synthesize the idea clearly_
the goal of humanity should be maximizing energy use per capita, this means every human should use 1000 times or 1M or 1B more energy than they do now.
this might sounds like a strange goal at first - for many reasons. first why would you want to increase consumption of something and not just get xyz outcome?
before making the point about why increasing consumption is good, let's create some common ground first.
i would argue money is fake - it is just a ledger for "the real currency"/raw input - energy. energy is definitionally wealth. this graph comes up a lot, as to say there are no low energy rich countries.
i think it is fairly intuitive to most how electricity spend is energy use but human labor is a type of energy spend, low efficiency high versatile energy spend. efficiency does matter but more energy spend is better than less even inefficiently - but yes the more leveraged energy spend you have the better
so the idea of energy being real wealth is fairly intuitive and well understood. what does this look like for an individual tho? rather than a country?
i would argue that your personal wealth is essentially how much energy you can leverage to make your life better - this can be an AC or using a plane or hiring a maid or at minimum just the food you eat.
so it makes sense that the more energy you consume, the richer you are and that poorer people should go up the energy consumption ladder to reach the level of wealth of the west. but the heart of this piece is that it is **actually really hard to use energy**.
if i told you to go spend 1GW tomorrow
1 GW is equivalent to about 1.3 million horsepower, but that's still a bit abstract. Here's a more tangible way to think about it: 1 GW is roughly the power produced by 2,000 Olympic-sized swimming pools filled with jet engines stacked on top of each other 😄 Or more practically: Imagine all the household power used by a city of about 750,000 homes running simultaneously - all their air conditioners, ovens, TVs, washing machines, and everything else running at once. That's approximately 1 GW of power consumption. Another way to visualize it: 1 GW is about the same amount of power as: Doc Brown's DeLorean time machine from "Back to the Future" needed (1.21 GW), A large nuclear reactor operating at full capacity, 3.125 million solar panels operating under ideal conditions, 431 utility-scale wind turbines running at full speed
it would be impossible for you to spend that, ChatGPT math but sounds reasonable, source
you would have to spend years inventing new technology to do so. one such technology might look like running your own supercomputer of power AIs to do extremely leveraged work? or using starship to commute across the globe Apparently starship consumes something like 120 GW to lift off? ChatGPT math again tho our boy sonnet contested that so not sure. source. this is all to say that to increase energy consumption to 1GWh/day for an average american would mean them using three orders of magnitude more energy than they do now. this would mean their lives would be unimaginably better than ours.
there is a sense in economics of chasing forever growth, and if that is possible - you are chasing growth for growth's sake as it makes your life better. i think a better way to frame this is chasing how much energy can we consume, and we know the limits to this - this is in some sense moving up the [Kardashev scale.](https://en.wikipedia.org/wiki/Kardashev_scale)
i think the framing that is lost in the Kardashev scale convo is how it is actually about increasing energy consumption per capita - enabling each person more leveraged energy use to make their lives better.
---
### Working Out Should Be Thought of More Like a Daily Commute—A Problem to Solve, Not Accept
*May 14, 2023*
**Tags**: technology
> obesity is an engineering problem not a willpower problem
> note: this is a reskin of an essay i wrote a couple of years back where i polished it up and reposted it after bring this new site up, the original was written by hand this was [vibes written](/writing/vibes-writing)
Going to the gym is **a relic of an outdated human operating system—one designed for a world that no longer exists.** We have all this technology, all this innovation, and yet millions of people still spend hours every week running in place like hamsters, all to counteract the basic fact that we have too much food. This is stupid. The future should be calorie-agnostic. Workouts should be for masochists and hobbyists—the rest of us should demand a better operating system.
### **We Are Malfunctioning Meat Computers**
Nothing about modern life is natural. Our bodies? 200,000 years of software built for survival, not leisure. We were designed to **panic about starvation,** not count macros. Our metabolism is still operating in caveman mode, screaming: **“Store the fat! Famine is coming!”** Meanwhile, famine is not coming. If anything, the opposite—an infinite buffet of dopamine-laced sludge, precision-engineered to hijack our taste buds and make us balloon up like foie gras geese.
A quick aside: ever wonder why cancer is such a big deal now? Because for most of history, **something else killed us first.** Same with obesity, diabetes—diseases of abundance that our bodies were never built to handle.
 ](https://prod-files-secure.s3.us-west-2.amazonaws.com/aa7b5714-a7d8-4439-9feb-4ad60c4a5ac3/6f1fa63a-80af-4969-bafc-a4faa0c225e0/image.png?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Content-Sha256=UNSIGNED-PAYLOAD&X-Amz-Credential=ASIAZI2LB466VFBP4HPW%2F20260907%2Fus-west-2%2Fs3%2Faws4_request&X-Amz-Date=20260907T215744Z&X-Amz-Expires=3600&X-Amz-Security-Token=IQoJb3JpZ2luX2VjEH0aCXVzLXdlc3QtMiJIMEYCIQD56XtxEUayn2NI0BWgEryvsNJqz8VbwZUvYnHRYkMGkgIhALZmkhGIOIKCmzCpNSfBx4iG6dToAgHcFrgEyrbng2oDKv8DCEYQABoMNjM3NDIzMTgzODA1IgyNwIERyDMn8YV8EZIq3AMMyX%2FN9cY%2B5IQW7REQHAcLXQrlarZt8mlWO40xqOmq6vS1wL13DCrINKBfD2goIDN0VlR7SQH%2FPfYv1ffDJWqD1sKloZO6Ns22fnc%2FrBDsiXeBCUyPGxHbMx1Yvj7rPUm7%2FwXfxNdCc98BVKUPPHdvk32CzR3Ktsqrz82Hcms44v%2FHhQWcgVwwEnkQ2wTEhx%2FKutAYiKfaj6W8ITCePsOQpvc6Lu%2Fjme1phisHVrLFEpr67epLoHo2KVzH5qoebQ1ldAAKXACYg7O7HuLxL2KyWG%2BAcQQV3s6jkCBUOHoTZXy%2BEH2p7rcMRzq%2BOBmO7Ex65mVzh7nlxqopuA%2BIvl2s13VX47MU%2FnB2A43JCoP647NkOmYcRa%2BADHMxZBIRvwq6uo02h9gdopt%2B16hEEjG089St1dUFDy473c3KmXVu%2BCpUwFav6dXFQOABZ%2BPsJ%2F3mohSlSrhmH6XuOwKfLoJFAXiIihDjBO48tX69RcjFJHslPOGEu5uc74MbuJRgHTJc3XdX29o%2FwGXWEyNtZonlot2sjeu2MalsQ2p4bOYuIZQvfKj%2Fce39hhU%2B3aIWYl7oMTpSiMjv%2B%2FCxd1yn36JWjWQXcDaBfLLW5XtQcRVI8DAPac5NlY2mFa5AXjCOvvzUBjqkAftm%2BRA6XuuKn%2FjOc%2FCCOIY6Vg%2Fs8sX8jeowE6HDWnecv369%2Fh3XaOtnNAXiDg9tZRC76d3dnyA6LY%2F%2FTqou9H5xqtaX8Pnj9xjjSVsVsrxMhVTPOMNtCOhu2TYFtaobf6uLNzzKf%2BGwinm9bKAQUXM3dDGPgwJg3RD0uHo78OiP1feYbD%2BNIZ853qL%2FvAqK5iMYAPFIC5S7UOY6Pfi2oyEW1FKG&X-Amz-Signature=cf39d969d34d94f8ff4f6a20ef3296d641d45cfc49d4a1878ff517aaa9659127&X-Amz-SignedHeaders=host&x-amz-checksum-mode=ENABLED&x-id=GetObject)
### **Stop Blaming People, Blame the OS**
The dumbest possible take on obesity is **“People are lazy.”** That’s like blaming a computer for running buggy software. We don’t need better *willpower*, we need better engineering.
**We should be solving this with technology, not discipline.** Imagine a world where you can eat anything, whenever you want, and remain **perfectly healthy**—not because you’ve “optimized your macros” like a psycho, but because **science has already solved it for you.**
Palmer Luckey has framed this problem correctly: we should not be trying to guilt or force people into dieting and exercising. That’s a brute-force approach to a problem that requires real technological intervention. If we can solve this at a fundamental level, why shouldn’t we?
### **The Gym Is Just a Commute for Your Body**
Nobody actually *likes* commuting. It’s a tax on your time. Sure, some people love driving, but the majority? **A colossal waste of man-hours.** This is how we should think about the gym. **Most people don’t actually enjoy lifting heavy things and putting them down—they do it because they have to.**
Self-driving cars will eliminate commutes, and **we should be using the same mindset for working out.** The goal is not to “make fitness fun,” the goal is to make it **irrelevant.** You should **be fit by default.** The only people who should be going to the gym are the ones who genuinely love pain and suffering (which, to be fair, is a respectable hobby). The rest of us should have technology do the work.
### **The Future: Food Without Consequence**
We’re thinking too small. “Healthy eating” is a brute-force solution. **We need engineered food—food that tastes exactly like a double bacon cheeseburger but has zero impact on your body.**
This isn’t impossible. We already bioengineer everything from corn to lab-grown meat. Why not take it further? Why are we still accepting **caloric reality** like peasants? We should be eating **synthetic, calorie-perfect, zero-consequence food.** Not *fake* food—real food, perfected.
### **We Have to Take the Blame—And Fix It**
The fact that this problem is still unsolved is a failure of **us, the builders.** It’s easy to point fingers at the food industry, the fitness industry, or bad personal habits, but the truth is, **if we were serious about fixing this, we would have by now.**
Technologists, engineers, biohackers—we should own this failure. We created a world where calorie-rich food is abundant but calorie-burning is optional. We can’t just shrug and tell people to “try harder.” We need to be thinking bigger.
This is something that has been circulating in my brain for a while. It feels like a lingering piece of an idea maze—one I haven't fully mapped out yet, but I know is important. If you're working in this space, I would love to chat, to jam, to be helpful, to collaborate. Maybe this is something I’ll work on myself in the future. But someone needs to build this. **This should exist.**
The future isn’t about fitness—it’s about **making fitness obsolete.** Imagine telling someone from 2100 that we used to spend hours every week *manually burning calories* because we hadn’t figured out a better way yet. They’d think we were **insane.**
### **Final Note: If You Crack This, Don’t Go to Mars**
If you get the biotech upgrade that lets you eat everything and stay shredded, **do not move to Mars.** That’s not going to be a post-scarcity society. If things go south, **you’re going to want your original metabolism back.**
But Earth? Here, in the land of abundance? We have **no excuse** for still doing things the old way. Fix this, or future generations will laugh at us.
The gym is a tax on human failure. It’s time to zero it out.
---
### data moats are not real
*August 3, 2025*
**Tags**: ai, startups
> why betting on proprietary data as your competitive advantage is an anti-bitter lesson argument
import { NoteReference } from "~/components/NoteReference";
_been thinking about how people are positioning data moats as competitive advantages in ai, and i'm increasingly convinced they're making an anti-bitter lesson argument - writing this down to clarify my thinking_
## the anti-bitter lesson
here's what i keep seeing: companies that think specific data moats exist are essentially betting against the bitter lesson. they believe their proprietary datasets, their special fine-tuning, their unique domain knowledge encoded in data will be their durable advantage.
rich sutton's [bitter lesson](http://www.incompleteideas.net/IncIdeas/BitterLesson.html) teaches us that "general methods that leverage computation are ultimately the most effective, and by a large margin." for 70 years, we've watched domain-specific approaches get demolished by whoever just threw more compute at the problem.
## scale wins again
take gpt-4-turbo and do extensive supervised fine-tuning on zig programming (which it's bad at). you might think you've created something valuable. but claude-4-sonnet will demolish your specially trained model just by being bigger and better at the base level—no special zig training needed.
this pattern plays out everywhere - the best chess engines don't have special chess data, they just have better search. the best image models don't have special art history databases, they just saw more of the internet
this isn't just about programming languages. in computer vision, we saw the same pattern—"early methods conceived of vision as searching for edges, or generalized cylinders, or in terms of SIFT features. but today all this is discarded." the winners didn't have better data; they had better ways to leverage compute.
## missing pieces
there are really two pieces that haven't been solved yet:
1. being a good base knowledge worker
2. some form of continual learning (as [dwarkesh points out](https://www.dwarkesh.com/p/timelines-june-2025) — models can't yet learn your preferences the way a human employee would)
both of these are extremely scale-prone. they're not defensible positions—they're just temporary limitations at current levels of intelligence.
it's not obvious to me what other specific jobs need tuning, but even those are likely scale-prone and/or won't need much training at the next oom of intelligence.
if gpt-5 or claude-5 is as big a jump as 3→4, most "specialized" models become instantly obsolete. the gap between general and specialized narrows exponentially with scale
## so what
this is both for how to think about moats while building companies in ai and how to choose problems while building companies.
if your entire thesis is "we have special data," you're making a bet against the trajectory of ai progress itself. better to think about problems that:
- leverage ai but aren't just ai wrappers
- solve coordination/trust/human problems that pure intelligence can't
- build network effects or switching costs unrelated to model performance
i could be very wrong here, but i don't think i am. scale has been undefeated. the bitter lesson always wins.
---
_evolving thoughts—would love to hear counterarguments or examples where data genuinely creates lasting advantage. find me at [hey@cryo.wtf](mailto:hey@cryo.wtf) or [book time here](https://cal.com/cryogenicplanet)._
this essay was somewhat [vibes written](/writing/vibes-writing)
---
### new axis of competition between startups and incumbents
*February 23, 2025*
**Tags**: ai
> how do startups compete with test time compute
> i don't have the answers and i am thinking out loud, this should sorta be read with my [really building for agi post](https://cryo.wtf/writing/building-for-agi-future)
for years, the battle between startups and incumbents was a fight between **bloat and distribution**. could startups stay lean enough to outmaneuver the giants? could incumbents leverage distribution to crush their smaller, faster-moving rivals? that was the equation. but now, there’s a third axis—**test-time compute**—and it changes everything.
### what is test-time compute and why it matters
test-time compute is running a model for longer, allowing it to think more. this is what powers advanced reasoning models—your O-series models, your chain-of-thought prompting, tree-of-thought exploration, monte carlo tree search (MCTS). it’s about generating better answers not just by being smarter initially, but by iterating, refining, and reasoning through a problem dynamically. every additional second of test-time compute gives an ai system the ability to re-evaluate, simulate different possibilities, and optimize outcomes, making intelligence itself a function of budget.
in business, many decisions come down to simple capex ROI math—if spending \$50 billion on compute increases conversion rates by 5% and that 5% drives \$60 billion in revenue, the decision is easy: just spend more. incumbents can turn on this loop and pour infinite money into refining intelligence-driven optimizations, whether it’s maximizing engagement on instagram, fine-tuning ad targeting, or optimizing search placement.
test-time compute becomes the ultimate strategic weapon—an unbounded ability to generate better answers by thinking harder and longer. and in an environment where intelligence can be brute-forced through spending, the company with the deepest pockets will always win.
this is why startups **must shift their definition of intelligence**. the old game of 'we'll just be better engineers' doesn't work when the real advantage comes from throwing infinite compute at a problem. startups need to be smart not in raw intelligence, but in picking **problems where intelligence itself isn't the currency of competition.**
### problem selection matters more than ever
picking the right problem has always been crucial in building a successful startup, but in the AI age, it's more important than ever. before, startups could win by building better software, designing more efficient systems, or optimizing workflows in ways incumbents couldn't match. but now, intelligence itself has become a commodity—one that can be scaled infinitely with compute.
this means that startups can no longer rely solely on being 'smarter' in the traditional sense. instead, they need to be smarter about where they apply intelligence. the key is to find **problems where the marginal returns to intelligence are low**—where throwing more compute at the problem doesn't yield significantly better results. this is where startups can still have an edge, not by outthinking in the raw sense, but by out-positioning incumbents through problem selection.
quick tangent on what is marginal returns to intelligence, this is straight from [dario amodei's excellent machines of loving grace](https://arc.net/l/quote/usalpydw)
> Economists often talk about “factors of production”: things like labor, land, and capital. The phrase “marginal returns to labor/land/capital” captures the idea that in a given situation, a given factor may or may not be the limiting one – for example, an air force needs both planes and pilots, and hiring more pilots doesn’t help much if you’re out of planes. I believe that in the AI age, we should be talking about the marginal returns to intelligence and trying to figure out what the other factors are that are complementary to intelligence and that become limiting factors when intelligence is very high. We are not used to thinking in this way—to asking “how much does being smarter help with this task, and on what timescale?”—but it seems like the right way to conceptualize a world with very powerful AI.
this is the fundamental insight: if intelligence alone determines success, the company that can afford to spend the most on compute wins. but if a problem has other constraints or bottlenecks that startups can solve they could win
### where startups could win
the key for startups is to identify **problems where intelligence is not the primary driver of success**. these include:
- problems where **intelligence isn’t the bottleneck** - maybe compliance, getting things done in the real world or **network effects (this feels like a big one).**
- markets where incumbents **don’t see value in deploying high test-time compute** because they don’t think it’s worth the spend.
- ??? - email me ideas [hey@cryo.wtf](mailto:hey@cryo.wtf)
if intelligence is just one input among many, then startups still have ways to compete but they should make sure to compete outside the axis of intelligence
*i am caveating small outcomes, this is always a thing a 200M business is rounding error for microsoft but might be great outcome for a founder, i am talking at scale*
this essay was [vibes written](/writing/vibes-writing)
---
### vibes writing
*February 22, 2025*
**Tags**: productivity, ai
> using ai to write fast and loose
## what is vibes writing?
vibes writing is writing without the hand-wringing, the overthinking, the self-important slow crawl to a perfect final draft. it’s writing how you think—fast, messy, in motion. this is what i’m doing right now. dictating ideas, riffing with ai, seeing what sticks.
it’s chat as a medium. it’s the opposite of painstaking craftsmanship. it’s *move fast, write things*, let them evolve.
## tools of the trade
vibes writing works because the tech finally makes it possible:
- **wispr flow** – speak, don’t type. typing is too slow for thoughts.
- **chatgpt projects + canvas**– where ideas go from scattered sparks to something real. the project pulls from my past work, guided by a custom system instruction that keeps the writing sharp, natural, and effortless
- **real-time feedback loops** – thoughts, iterated and refined, in motion.
it’s an engine for ideas. a way to turn a blank page into a written thing before you even know what happened.
## why vibes writing?
most writing feels like chiseling stone. slow. solitary. tedious. vibes writing? it’s like throwing clay on a wheel and shaping it as it spins. ideas don’t start perfect, but they start. and that’s the whole point.
- **speed** – the best ideas don’t wait.
- **flow** – writing should feel fun, not like homework.
- **authenticity** – write how you talk, not how you think you *should* talk.
## vibes written posts
some of my posts will be labeled “vibes written.” what does that mean? it means i didn’t sit there agonizing over every word. i spoke it, jammed on it, let it take shape in real-time. some posts will stay raw. others will turn into polished essays. but they all started here.
## what’s next?
no idea. vibes writing is an experiment. a work in progress. maybe it’s just how i write now. maybe it’s bigger than that. who knows? but that’s the fun of it. no overthinking. just vibes.
this essay was [vibes written](/writing/vibes-writing) :)
---
### rahul’s raycast guide
*February 10, 2025*
**Tags**: productivity
> how i use raycast
*it is interesting to me that i haven't written this before since i use raycast and shill it so much but this is an extension of my *[*macos guide*](/writing/macos-guide)* specifically for raycast*
high level, raycast is a spotlight or alfred ++ and does a lot of basic things really well. i will run these down quickly but the main point of raycast is how custom it can be
### basics
- launch apps - search launch apps fast, has good aliasing to rename apps imagine you want to call notion calendar cron or whatever else
- calculator - god send, its not just math but also unit conversion, time zone conversion and even date stuff
- file search - fast first search i have it bound to alias `f`
- clipboard history (massive can’t use a computer without this) - have it bound to `cmd` + `shift` + `c` even has images stored in clipboard and can search it is great
these are all inbuilt and amazing features that itself make raycast a must have the clipboard history is a god send
tho a lot of people cover this and really you should just play with this and explore
## using extensions and keyboard maxing your computer
this section i will talk about raycast’s extension store and keyboard maxing many if not all actions on your computer
as you can see i have lots of customised things here and its not that i only use the customised this or even that i use all of them but rather it is a good indication of how/where i use raycast will run through a few here and then later can show up truly custom things
going to just highlight my most used `aliases` and `shorcuts` as the tools i use
- bitwarden - my password manager, i love `sv` to retrieve passwords and `gp` to generate a password
- window management - i have rectangle and use that for dragging but love `lh` `rh` and `mm` as shortcuts to move windows around fast
- spotify - i really like `pp` as toggle play pause as sometimes native media play pause fucks up when there are many things playing
- `dw` - dictionary i use this to basically check spellings tbh
- `kp` - is kill port when there is something dumb on a port and want to get rid of it
- `ej` - eject disks
- `bt` - open up bluetooth connections (extension name toothpick)
- `od` - set which device is audio out
- `option + cmd + space` raycast emoji picker - is great
- `ss` - search screenshots (including with ocr on the text inside)
- `ry` - you can move raycast around so this resets it back
- `rp` - remove paywalls on newsites
- `snip` - not a huge snippet user, but use it to store general links
i am sure i am missing somethings and not all extensions get used all the time they just stack so when you need them they are there
## truly custom stuff
the main thing i love about raycast is it so damn easy to right extensions to do what you want
raycast has three sorta ways you can do this
- quicklinks
- raycast scripts
- raycast extensions
let me show you examples of each
whatsapp doesn’t let you text people without saving the contact but you can do [`wa.me/number`](http://wa.me/number) so i have a quicklink for that so `wa` `tab` `number` opens a wa chat with the person
another fun version of this is for github codesearch `csm` redirects to which search that org and you can imagine having one for each org
`https://github.com/search?q=org%3Ascalarhq{argument name="Argument"}&type=code`
tbh i don’t use these too much but when i do its handy, i used to use the `!p` and `gg` before i made [pplx-bangs](https://github.com/CryogenicPlanet/pplx-bangs)
### script commands
i only have a few here mostly around macos `space` management using a thing i built on top of `yabai` called [`fenster`](https://github.com/CryogenicPlanet/fenster) i will be the first to admit that it is quite jank and doesn’t fully make spaces great so i just mostly use window management and disregard spaces when i do use them this is a lifesaver tho
the `mem` command adds tools to #tools on my [mem.ai](http://mem.ai/) and this uses an `zx` script so all scripts don’t have to be bash
```javascript
#!/usr/bin/env zx
/*
# Required parameters:
# @raycast.schemaVersion 1
# @raycast.title New Tool
# @raycast.mode compact
# Optional parameters:
# @raycast.icon 🤖
# @raycast.argument1 { "type": "text", "placeholder": "Link" }
# @raycast.argument2 { "type": "text", "placeholder": "Body", "optional": true }
# @raycast.packageName Mem.ai
# Documentation:
# @raycast.description Add a new mem
# @raycast.author Rahul Tarak
# @raycast.authorURL https://github.com/CryogenicPlanet
*/
const arg1 = process.argv[3];
const arg2 = process.argv[4];
let content = `#tools #fromRaycast ${arg1} ${arg2}`;
const resp = await fetch("https://api.mem.ai/v0/mems", {
method: "POST",
headers: {
"Content-Type": "application/json",
Authorization: "ApiAccessToken ",
},
body: JSON.stringify({
content: content,
isRead: true,
isArchived: true,
}),
})
const data = await resp.json()
await $`echo ${data.url} | pbcopy`
console.log("Copied url to clipboard!")
```
### extensions
the real fun!
i have more custom extension than this but i think you can move them out of dev state such that they don’t show up here or something
a few of my favourites
`md` - let’s you control which display is main display from raycast built on top of this [https://github.com/CryogenicPlanet/hoveddisplay](https://github.com/CryogenicPlanet/hoveddisplay)
`td` - spotify devices let you change which device music is playing on from raycast
- `sidenotes` - gives read + write functionality to [https://www.apptorium.com/sidenotes](https://www.apptorium.com/sidenotes) from raycast
- `wiz` and `mirale` are to control my lights and ac from raycast so like smart-home stuff, they are for my smarthome lights in india so don’t use them when i’m not there
- `GPUs` where to have a menu bar of out gpus on runpod and turn them off, had added some other providers after too
- `email domain extactor` is my fav kind of raycast extension, you are doing something repetitive like going through a list and want to get domain to email you write an extension in 5 mins and it saves you a lot of time doing that task - you may never use it again, it was bound to `option` `l`
the other extensions here are like trying to build a semantic search index and memory thingy to search my computer, didn’t really finish that
if you notice for a bunch of these quicklinks and extensions they are shared to `1202` that is just my friends and sometimes we share these internal tool inside an org too and it is great for building quick internal tools
raycast is always adding new stuff and i would just recommend playing, there are definitely tons of things i use that i didn’t talk about and new things i am trying like their new hyperkey support
for fast opening apps bound to `caplock` my hyperkey
if you are using raycast in fun ways let me know [hey@cryo.wtf](mailto:hey@cryo.wtf)
---
### Rahul’s Mac Productivity Suite
*December 19, 2024*
**Tags**: productivity
> a guide to making the most of your mac
import { NoteReference } from "~/components/NoteReference";
_this is a refresh of a guide i wrote a few years ago_
## Battle Tested
1. [Raycast](https://www.raycast.com/) - the best way to use your mac
raycast does so much for you, i would just play with it and see what it can do. i highly recommend the clipboard manager (i have it set to `cmd + shift + c`). raycast is also awesome to build custom extensions, would highly recommend this too.
i could go on for days on great extensions here, but would just say explore and you will find so much.
2. Window management - unfortunately macos natively sorta sucks at this
three options:
- [Rectangle](https://rectangleapp.com/) - open source pretty good!
- [Magnet](https://apps.apple.com/us/app/magnet-window-manager/id441258766) - i use this, mostly cause i paid before i found rectangle
- use raycast itself? i do this too, but i use it differently and would recommend one of the above for dragging bounding boxes which raycast does not support
3. [Warp.dev](https://www.warp.dev/) - personally really like it as a terminal, heard great things about [ghostty](https://mitchellh.com/ghostty) haven't tried yet. personally like an `oh my zsh` config with `starship` here but anything sorta works. would recommend adding the `zsh-autosuggestions` plugin
4. for the love of god use [cursor](https://www.cursor.com/), i don't think this needs to be said! setup `c .` to open cursor in the current directory.
not the biggest `vim` user but would recommend using, and have relative lines for sure and add the following to your vscode settings, makes copy paste work
```json
{
"vim.smartRelativeLine": true,
"vim.handleKeys": {
"": false,
"": false,
"": false,
"": false
}
}
```
5. [Arc](https://arc.net/gift/961c1d38) - it is great!
6. the chatgpt app is good, claude desktop kinda sucks - haven't tried perplexity desktop.
7. cron/notion calendar is decent calendar, i like it! don't love it but does the job
8. screenshots: macos default is good, but cleanshot x is amazing
setapp is a single subscription to many pro apps. here are all my apps
but also shottr is free alternative that is decent
9. notch: please get bartender (through setapp) or [dozer](https://github.com/Mortennn/Dozer) free alternative
10. stats: istat menus is good (through setapp) or [stats](https://github.com/exelban/stats) is free alternative
11. mail: superhuman is amazing, expensive but amazing. looking forward to trying notion mail.
12. **[beeper!](https://www.beeper.com/)** this app is so fucking good - it connects all your messaging apps into one and there is a new version too [beta here](https://www.beeper.com/beta)
13. search: not as much a mac app perse but still fun search tool i use is [`pplx-bangs`](https://github.com/CryogenicPlanet/pplx-bangs) which let's me route between `google`, `perplexity` and `exa`. you can customize defaults and routers for yourself!
for writing i use notion, but honestly anything works.
## Extras/Config
- please for the love of god, turn this off - it is the worst thing ever
- turn this off too, this for some dumb reason takes port 5000 and keeps restarting very annoying.
here is a more clear view of all my setapp apps, would really recommend mission control plus and popclip, apart from the ones mentioned.
- setup `sudo` to use touch id, don't remember the exact steps ask chatgpt but you can easily do this - it is great!
- drag your download folder to your dock and set it to `fan` and sorted by date modified, it is very good to drag stuff from there
_will update this guide as i remember more things_
---
## Ideas
a lot, maybe most **“venture” backed founders** don’t actually want to make **venture scale companies** - a lot of people talk about how venture went from this niche boutique business to this massive thing but i think this is miss understood. so many times i’ve heard i want to make b2b or devtool but then stay stripe then say n of 1, can’t replicate or you know company’s founder made money who cares if it went nowhere
---
i am generally a big memetic fitness guy and just found this interesting as a counteer from qwern - memetic fitness in ideas are weird, might not be as strong to use them as a judge of fitness of ideas like biological ideas [https://arc.net/l/quote/ymqwjqkm](https://arc.net/l/quote/ymqwjqkm)
> my counter would be: there are some ideas that die by just having bad TFR and actually go extinct causing at least at some level a real fitness function here but this is a weak selection pressure and acts slowly but can be exponential
---
this is the longest time in history we have had peace → same as nothing is changing???
> not the end of history we have solved our problem and have deterance
> but rather both the peace and the stagnation create these loop of each other?
---
---
there is such a thing as **value**, it is definitely** proportional to money but money is a bad proxy **maybe value is energy in the abstract but this quest for what is really valuable in the world is an important one that i don’t think most people spend much time thinking about
---
i believe in some capacity if you have the ability to do great work, then it is your “moral imperative” to do so - to do the hard things and reshape the world
---
there aren’t 180-190 sovereign countries, i’ve been noodling in my head that there about 20-25, that can make truly independent decisions some examples india (can say no to russian oil sanctions or fuck off to canada), saudi, israel, obviously the us, china and russia - definitely not europe
> a samo burja claim is that there are even less maybe 6-7 it’s not just ability to make these independently decisions today but the sorta land mass plus population plus industrial base and such that creates very few truly sovereign blocks
> a second point here, the stated in the empire aren’t powerless to its leadership obviously the 50 states in the union have direct power but the numerous other states in pax american have influence and power
---
very half baked but essentially there seems to be some kind of cost for dynamism, for growth. a cost to social stability or at least to the status quo
> dynamism, progress, growth etc all seem to be good things - things everyone should want but there seems to be some cost to bear that many large economies are unwilling to
> this cost seems to be a sorta prisoner’s dilemma type cost, where someone has to pay it otherwise the world suffer badly but others get to cheap by leaching off the person paying the cost
---
it seems like last cycle of software companies were very end state with their technologies and thus very high craft very well built software, but it seems like lot of ai companies today are not at all that; in some distinct way it seems the quality of software has deteriorated, i wonder if this is the language models writing code or the founders this technology is choosing for
---
[https://worksinprogress.co/issue/the-housing-theory-of-everything/](https://worksinprogress.co/issue/the-housing-theory-of-everything/) great piece, one thing to pull out here is very weirdly interesting to me is how housing costs suppress productivity but pricing out less productive employees making the more productive employees do their jobs; like probably lot of tech people should have in person EAs or Cleaning Staff but don’t cause it is too expensive making them do the less productive task
> this was on a pod but an interesting pull out from the same team was how as people don’t bunch up as much anymore there aren’t 8 people living in one dwelling the housing shortage in cities like toronto is much more acute like they are missing something like 40% housing stock just for the people currently living in toronto
---
## Musings
# Rahul Tarak - Musings
Short notes on things I am reading, watching, and thinking about.
## Reading
*articles, papers, posts, and rabbit holes*
### undefined
*Aug 4, 2026*
### undefined
*Jul 22, 2026*
- [oxc.rs](https://oxc.rs/blog/2026-07-22-type-aware-linting-stable)
### undefined
*Jul 16, 2026*
- [sierra.ai/blog/horizon](https://sierra.ai/blog/horizon)
### undefined
*Jul 8, 2026*
- [bun in rust](https://bun.com/blog/bun-in-rust)
### undefined
*Jul 8, 2026*
- [typescript native port](https://x.com/ahejlsberg/status/2074899956511760806?s=20)
### undefined
*Jul 8, 2026*
- [gpt live](https://openai.com/index/introducing-gpt-live/)
### undefined
*Jul 6, 2026*
- [july 2026 workspace](https://transformer-circuits.pub/2026/workspace/index.html)
### undefined
*Jul 3, 2026*
- [rift](https://github.com/anomalyco/rift)
### undefined
*Jun 27, 2026*
- [x post](https://x.com/richzou/status/2070681443379130832?s=46)
### undefined
*Jun 26, 2026*
- [realtime subagents](https://www.inceptionlabs.ai/blog/rise-of-realtime-subagents)
### undefined
*Jun 19, 2026*
- [paper](https://arxiv.org/pdf/2401.05566)
### undefined
*Jun 11, 2026*
### undefined
*Jun 2, 2026*
### undefined
*Jun 2, 2026*
### undefined
*May 31, 2026*
### undefined
*May 30, 2026*
- [github.com](https://github.com/ComposioHQ/chorm)
### undefined
*May 14, 2026*
### undefined
*May 14, 2026*
### undefined
*May 13, 2026*
### undefined
*May 7, 2026*
### undefined
*Apr 29, 2026*
- [x.com](https://x.com/eriktorenberg/status/2049194746883752072?s=46)
### undefined
*Apr 25, 2026*
### undefined
*Apr 17, 2026*
### undefined
*Apr 17, 2026*
### undefined
*Apr 17, 2026*
### undefined
*Apr 7, 2026*
- [ntrs.nasa.gov](https://ntrs.nasa.gov/api/citations/19720005243/downloads/19720005243.pdf)
### undefined
*Mar 29, 2026*
- [x.com](https://x.com/_chenglou/status/2037715226838343871?s=46)
### undefined
*Mar 29, 2026*
- [x.com](https://x.com/_chenglou/status/2037715519277760531?s=46)
### undefined
*Mar 29, 2026*
- [github.com](https://github.com/chenglou/pretext)
### undefined
*Mar 20, 2026*
- [rivet.dev](https://rivet.dev/)
### undefined
*Mar 17, 2026*
- [github.com](https://github.com/theredsix/agent-browser-protocol)
### undefined
*Mar 16, 2026*
### undefined
*Mar 12, 2026*
- [agents.ramp.com](https://agents.ramp.com/cards)
### anthropic jobs data
*Mar 6, 2026*
- [pdf](https://cdn.sanity.io/files/4zrzovbb/website/dc7bcd0224644fce97cecb7f9e68dcd8434b35f1.pdf)
### harness engineering
*Mar 4, 2026*
- [read](https://openai.com/index/harness-engineering/)
### why europe lacks tesla
*Mar 3, 2026*
- [works in progress](https://worksinprogress.co/issue/why-europe-doesnt-have-a-tesla/)
### recent reading
*Mar 3, 2026*
recent reading.
- [stratechery](https://x.com/stratechery/status/2028425096054931921?s=46)
- [dean ball](https://x.com/deanwball/status/2028464782622195992?s=46)
- [eoghan mccabe](https://x.com/eoghan/status/2028522852044206258?s=46)
### saaspocalypse
*Feb 20, 2026*
the one in which sarah guo and elad gil talk through the saaspocalypse.
- [x thread](https://x.com/saranormous/status/2024581299491787219?s=46)
### stripe minions
*Feb 20, 2026*
- [stripe](https://stripe.dev/blog/minions-stripes-one-shot-end-to-end-coding-agents-part-2)
### ramp ai index
*Feb 12, 2026*
very interesting.
- [x thread](https://x.com/arakharazian/status/2021637704039993437?s=46)
### undefined
*Jan 26, 2026*
### claude in excel
*Jan 24, 2026*
- [x post](https://x.com/claudeai/status/2014834616889475508?s=46)
### codex agent loop
*Jan 24, 2026*
- [read](https://openai.com/index/unrolling-the-codex-agent-loop/)
### undefined
*Jan 21, 2026*
### undefined
*Jan 20, 2026*
- [x.com](https://x.com/linear/status/2013643099147248054?s=46)
### undefined
*Jan 20, 2026*
- [nadia.xyz](https://nadia.xyz/basic)
- [johnsalvatier.org](http://johnsalvatier.org/blog/2017/reality-has-a-surprising-amount-of-detail)
### undefined
*Jan 20, 2026*
- [vercel.com](https://vercel.com/docs/microfrontends)
### undefined
*Jan 16, 2026*
### energy of tastemakers
*Jan 13, 2026*
i love this description, so much of what i do and the pain with it.
- [essay](https://hvpandya.com/energy-of-the-tastemakers)
### undefined
*Jan 8, 2026*
- [x.com](https://x.com/tbpn/status/2008967307113189385)
### undefined
*Jan 6, 2026*
- [cs.cmu.edu](https://www.cs.cmu.edu/~pavlo/blog/2026/01/2025-databases-retrospective.html)
- [x.com](https://x.com/swyx/status/2008181221587906688)
### undefined
*Jan 2, 2026*
- [x.com](https://x.com/rahulgs/status/2006090208823910573)
### undefined
*Jan 2, 2026*
- [x.com](https://x.com/bcherny/status/2007179832300581177)
### undefined
*Jan 2, 2026*
- [danwang.co](https://danwang.co/2025-letter/)
### physical intelligence
*Dec 24, 2025*
insane. would also recommend reading their original paper, super interesting.
- [x post](https://x.com/physical_int/status/2003161637734518985)
- [openpi](https://www.physicalintelligence.company/blog/openpi)
### chain of thought monitorability
*Dec 19, 2025*
bookmarking for myself for later.
- [openai](https://openai.com/index/evaluating-chain-of-thought-monitorability/)
### undefined
*Dec 19, 2025*
### undefined
*Dec 18, 2025*
- [x.com](https://x.com/tbpn/status/2001439909530472761)
### undefined
*Dec 18, 2025*
- [x.com](https://x.com/OpenAIDevs/status/2001419749016899868)
### undefined
*Dec 17, 2025*
- [x.com](https://x.com/mntruell/status/2001417450542576107?s=46)
### factorio learning env
*Dec 17, 2025*
forgot to play with this, but it was my favorite paper from neurips.
- [github](https://github.com/JackHopkins/factorio-learning-environment)
### langwatch scenario
*Dec 17, 2025*
- [github](https://github.com/langwatch/scenario)
### undefined
*Dec 17, 2025*
### works in progress run
*Dec 17, 2025*
i've loved works in progress the last few months. they are publishing great work.
- [downzoning](https://worksinprogress.co/issue/the-great-downzoning/)
- [housing theory](https://worksinprogress.co/issue/the-housing-theory-of-everything/)
- [baby bust](https://youtu.be/sI0h6-wec98?si=y87kM1CBba6X4FZi)
### undefined
*Dec 17, 2025*
- [x.com](https://x.com/lulumeservey/status/1998139088696963094?s=46)
### undefined
*Dec 17, 2025*
- [delian.io](https://delian.io/lessons-5)
## Thinking
*half-formed takes that are not essays yet*
### undefined
*Jul 10, 2026*
### undefined
*Feb 9, 2026*
### on dairo's writing
*Jan 26, 2026*
dairo is such a good writer, the articulation and choice of words is so good.
## Watching
*talks, demos, and things worth sitting with*
### undefined
*Aug 6, 2026*
- [youtube.com](https://www.youtube.com/watch?v=87DyyMV0kCY)
### undefined
*Jul 29, 2026*
- [x.com](https://x.com/patrick_oshag/status/2082073587142234374/video/1?s=46)
### undefined
*Jul 24, 2026*
- [youtu.be](https://youtu.be/z6gH_v0buUc?is=SiCU5bL5XJJQJUfe)
### undefined
*May 13, 2026*
- [x.com](https://x.com/patrick_oshag/status/2054532117410054252/video/1?s=46)
### undefined
*May 1, 2026*
- [youtu.be](https://youtu.be/I-ldITom1cg?si=kJgnhKgwhAJBgNQo)
### undefined
*Apr 9, 2026*
- [youtu.be](https://youtu.be/dvt_74kV-RM?si=eEe_Vw_wR6n1-9sr)
### undefined
*Mar 10, 2026*
- [futureblind.com](https://futureblind.com/p/atoms-are-cheap-process-is-pricey)
- [youtu.be](https://youtu.be/owZ6xh_gOdY?si=e4ys4WBHldCBPyws)
### undefined
*Jan 23, 2026*
### gpt-5.2 browser demo
*Jan 14, 2026*
wow. this is crazy.
- [part one](https://x.com/mntruell/status/2011562190286045552)
- [part two](https://x.com/mntruell/status/2011562192601305091)
### brie wolfson on craft
*Jan 6, 2026*
i think brie articulates company culture, craft, and taste things very well.
- [watch](https://www.youtube.com/watch?v=GqVv_e52Z2Y)
### red bull founder
*Dec 17, 2025*
i knew very little about the founder of red bull and this was just very interesting.
- [watch](https://youtu.be/nrmjb5PC0ms?si=-PXLeG_N-PJ363-H)
