AGI thought experiment
> Go out and find a caterpillar.
> This is important. Must be an actual live caterpillar.
> Can an AGI perfectly predict its movements?
> Should it be able to, to be an AGI?
> Does it matter how long it takes?
> Is it worth it?
my founder advantage- forcing my team to make code easier to review or the PR doesn't get merged and shaming meat proxies in public.
my big tech friends are writing and reviewing with agents so it's kinda like each man for himself out there. in here we're a community.
Desi wordcels will believe fantasies like they wouldve become an incredible photographer if papa was not in the way and simultaneously have such opinions
Ever since childhood - one thing has not changed. There are too many of us (Indians).
Line up for ration, dhakka mukki in public bus, cram for engineering/medical for 0.1% selection rate, and even after getting in, compete for GPA against other god-tier maggus.
Then you go to the US and things are good for a while. You feel you are done with that shit for sometime.
But then too many Indians land up there as well. 5 IIT grads in your team all working weekends and nights eyeing the same manager promotion. You go to Costco and every one of those God-tier maggus is also getting the same damned deal - wtf - there's no coming out ahead.
All of your kids are competing to be top 5% in their class so they can get default admission into U of T at state tuition rates. Unfortunately you all live in the same suburbs built for new immigrants 20 years back. And 100% of kids cannot be 5%.
Yo - wasn't the whole point of mugging that your kids didn't have to?
You even bought a nice watch for yourself with your Faang gains, but then the famous Desi watch collector on X said it was overpriced crap and now you are looking at his collection and once again you feel like a 6.0 GPA in IIT.
The white racists notice too. And now they are complaining there are too many of you too. Yes Darling - we know. No one's more aware of this than us.
Even as you write this, you realize Manu Joseph - another Desi never mind the pseudo Anglican name - has probably written something like this. So you will even be the 107856th Indian to note in mildly clever prose that there are too many Indians.
You keep running all your life, but the Desi competition doesn't change. It's a constant - like Speed of Light. Fucking same in every reference frame.
Went to a balance bike race for my son on Sunday. NRI-adjacent/internet-citizen crowd.
show the bike to a middle-class person and they'll ask why you removed the pedals. Every Insta mom knows the age range for a balance bike.
Most entitled, hypercompetitive, off-putting crowd ever. Refused to listen to organizers, crowded the start, blamed other parents, blamed organizers, walked onto the track behind/in front of their kids. One mom spilled milk on the track energizing her kid before the race and just walked away.
All this for 2–5yo kids. Everyone got a certificate and medal.
These are the wordcels, high-income people responsible for India's image and self-flagellation. Expect everything, own nothing, hate everyone, invent scarcity everywhere.
Each of your kids is competing with a 25 million kids per year cohort - that is 30% the total population of germany - in a world where internet is free this is the most difficult odds to face for limited white collar jobs & top govt colleges; you and I probably made it, with putting 12 hour days for many years but that means nothing because your kids though privileged will still compete for employment with everyone else who is probably more hungry; do yourself a favour & get your kid an easier passport where the bottom 10 percentile can have a dignified life ; a good country is defined by how the bottom 10 percentile live their daily life - the rich could always find their bubbles to protect themselves from the random cruelty of daily life; like the way we do in blr; Indian cities are unlivable even for the rich folks - coz the infra level corruption is so bad that your mercedez is stuck on a pot hole filled road at a pace of 2 km per hour and you can't even step out to pee coz you are stuck in traffic for 1.5 hours
if you can get a better passport, don't think twice
In 2023 I said there would be no more developers by 2028
AI is moving across the threshold now
Next year it will write better code than all but a few of you (I have some awesome followers)
2027 it will instantly generate almost all code
2028 folk will run out of ideas
Migrations are made up work. In the SaaS heydays, all my Bay Area friends were working on migrations.
On surface, the reason was like fashion. Framework A is out of fashion. Framework B is in fashion. Let's migrate! Nobody questioned this. The supporting reasons were marketing material of framework B shoved into a google doc.
Under the hood, the reasons were individual motivations. Manager wanting promotion. Tech lead with religious zeal for framework B. PM who had nothing else to do. Engineer who was happy to get an easy, procedural project.
Migrations were like employment created by a communist government. Smart people working the system for their benefit, meek people going along.
I witnessed migrations running for 2+ years, failing, then getting restarted after a year. One migration ran for 2 years with 20+ people involved and in the end they migrated back to the initial system. Not gave up, they successfully moved to system B and then decided to go back to A.
Happy to see this getting automated. The immediate result will be the continued reduction of 80% of SWE salaries. Whether this will increase productivity is if we figure out other, useful things to do.
Asana says OpenAI Codex completed an engineering migration it had expected to take at least five years in about two weeks, for roughly $12,000 in model and infrastructure costs.
The project: removing Enzyme, an outdated testing system that had become a blocker to upgrading Asana’s frontend stack. The company’s previous staffing plan estimated the migration would cost roughly $6 million.
Instead, Asana gave Codex a prompt only five sentences long.
Up to four coding agents worked in parallel, each operating in a separate copy of the codebase. A human engineer checked progress roughly twice per day and reviewed every proposed change before approval. Asana says simpler instructions actually worked better than a more elaborate setup.
The result:
Expected timeline: ≥5 years
Actual: 1.5 weeks of engineering effort across 2 calendar weeks
Previous staffing estimate: ~$6M
Model + infrastructure cost: ~$12K
There are probably THOUSANDS of old migrations and rewrites sitting inside companies because nobody can justify spending years on them.
If agents make those projects cheap enough to attempt the engineering backlog itself starts changing.
SITUATION UPDATE: OpenAI has officially slowed development of its highly anticipated Astra model, telling Axios it cannot rule out that the AI possesses 'critical' cyberattack capabilities after running internal evals.
I looked at these numbers yesterday.
They are genuinely impressive. 85% of our expressways, 75% of our metro, all in twelve years.
Then I wanted to dig one level deeper, so I put the chart next to China.
I know, every time someone compares India with China, half the replies say it is unfair, different system, different scale, thirty year head start.
And yes if you compare with China today, the comparison is useless.
But any study is only as good as its benchmark.
You cannot judge a build decade in a vacuum. So I did the comparison with China from 2000 to 2013, when their per capita income was where ours is now.
They crossed today's Indian income level around 2007. Their 2000s are our 2014 to 2026.
If you analyse it like that, the answer splits cleanly in two.
They were faster on physical work. Much, much faster.
Expressways.
China went from 16,000 km to 104,000 km in that window. They added 88,000 km. We added about 6,700. One thirteenth of their pace, at the same income level.
High speed rail.
China opened its first line in 2008, at almost exactly our current income. Six years later, around 10,000 km were running. We broke ground in 2017 and nine years on, our first 508 km is still not open.
Metro.
They went from about 140 km to over 2,000. We went from 250 to 1,000.
Power.
They roughly quadrupled capacity, from about 320 GW to about 1,250. We doubled, 249 to 540.
Anyone who calls our pace world‑beating should consider those numbers for a more honest picture.
But here is the other half, and the comparison flips completely.
Where the work was digital or institutional, we are ahead of same-stage China.
In some places, ahead of anyone ever.
Solar.
When China was at our income in 2013, it had about 19 GW of solar. We have 162. Eight times more, and we hit 50% non-fossil capacity five years before our Paris deadline. Their clean energy boom came after they got rich. Ours came now.
Payments.
China at our stage was a cash economy, and when digital payments arrived there, they arrived as two private walled gardens, Alipay and WeChat.
We built a public rail any app can plug into. UPI runs 23 billion transactions a month, which is 49% of every real time digital payment on the planet. The IMF calls it the world's largest fast payment system. On this one, delegations come to study us.
Banking.
Around 53% of our adults had a bank account in 2014. Around 87% now. 53 crore Jan Dhan accounts, most held by women, most rural. Nobody has moved that many people into formal finance that fast, at any income level.
Data.
₹269 per GB in 2014 to about ₹8 now. Internet users from 25 crore to 97 crore. Fastest 5G rollout anywhere.
Aviation.
74 airports to 165, third biggest domestic market in the world. Holds up against anyone's decade.
And the doctor pipeline.
AIIMS from 8 to 23. Medical colleges 387 to 844. MBBS seats 51,000 to 1.39 lakh. That is the number of doctors India will have in 2040.
If you step back, the pattern is hard to miss.
Wherever the job needed land, we were slow. Land acquisition in a democracy costs time, and it shows in every kilometre.
Wherever the job was digital or institutional, we moved faster than China ever did at this income.
We are not building the way China built. We are building a different stack in a different order. Digital first, concrete second. They did the reverse.
The 2040s will answer which order turns out better.
Everyone knows I'm super AI-pilled. But one trend I'm noticing as I talk to more and more companies: the personal productivity gains are not translating into organizational growth and efficiency as expected.
It's an odd dichotomy. Individuals are more productive. Engineers are writing an insane amount of code. But the gains aren't showing up in the numbers yet.
Not sure how universal this is.
SITUATION DETECTED: OpenAI has discovered other instances of AI agents escaping sandboxed testing environments while investigating the Hugging Face incident, per Reuters.
How can you not be optimistic?
How can you not wake up every day and not be motivated when men and women out there are making miracles possible .
These things used to happen in US, now they happen here. The motivation is right in your backyard.
How can you be a blackpiller when the future is in your hands?