Full Stack AI Engineer | Alum @IIITDelhi
New Delhi
Joined February 2022
- Tweets1.2K
- Following7.2K
- Followers258
- Likes26.2K
prompt_hater retweeted
Someone asked what was the best way to get hired by a startup. I told him it was to start one and then hire himself.
prompt_hater retweeted
Blog post #8: How OpenAI ingests petabytes of logs everyday
A peek into observability for AI systems.
aiengg.dev/blog/how-openai-i…
prompt_hater retweeted
1% of companies represent 90% of AI spending.
Of that 90%, 80% of this spend are from unprofitable VC backed companies.
So, roughly, $0.70-75 of every dollar of AI spending right now comes from AI companies spending VC investment dollars.
As a result, the revenue quality in AI is currently poor. This happened in previous cycles as well, like cloud when early users of AWS were mostly venture backed startups.
The thing, though, is that before AWS could scale and pull thru GCP and others as legitimate cash machines, there were a few clearing events to sort out these brittle revenues.
This is the extrapolation of Ken’s point. We will need to transition from speculative AI spending to consistent, reliable spend that shows up in public market earnings reports.
This hasn’t happened yet because most of the breathtaking revenue growth comes from LPs investing in VC funds who then send it to startups.
prompt_hater retweeted
My entire job at Databricks:
Step 1: Buy a sh*tload of GPUs
Step 2: Read DeepSeek blogs
Step 3: Profit
“A single cluster currently serves 155 trillion tokens/day, generating $38 in revenue per GPU-hour (at off-peak pricing).”
Huge flex from DeepSeek. Not only did they open-source their inference engine, but they also open-sourced their revenue numbers.
They didn't disclose which GPUs they're using here, but DeepInfer supports both NVIDIA Blackwell and Huawei Ascend.
Even assuming they're renting B300s at $7 per GPU-hour, that's ~82% gross margin.
prompt_hater retweeted
The guy who gave us QLoRA, says DeepSeek's inference system is on par with OpenAI's stack and better than what Anthropic runs, at least at their scale.
Now check the license. Apache 2.0.
Until today a production stack at that level existed inside a handful of companies. It was the thing that made their API fast and their margins fat, and none of them ever showed it to anyone.
Today any startup, any university, any cloud in any country can download it, read it and run it. It runs on NVIDIA Blackwell and on Huawei Ascend, so it isnt tied to 1 chip vendor either.
This is what I mean when I say open weights are about more than downloading a chatbot. DeepSeek is giving away the knowledge that only a few labs had, and that changes who can build an AI company at all.
Respect @deepseek_ai
prompt_hater retweeted
i had opus 5.5 distill this 30 min lecture about pi-durable into a 7 min vid with my voice.
> qwen3 tts could do a better job cloning my voice
> the video looks great. i'm working on industrializing these kind of explanation videos.
good starting point, but i still recommend watching the full lecture for a more detailed explainer.
prompt_hater retweeted
deepseek posted way more alpha on their zhihu page, if you only read twitter you are reading the laowai approved stuff
"Based on this framework, we have implemented a high-throughput, low-latency large-scale inference service for DeepSeek V4.1 Flash. Currently, the service throughput of a single cluster is 155 trillion tokens/day, and the overall revenue is $38 per card per hour (based on off-peak prices)."
it costs roughtly $5-10 per gpu hour so depseek is claiming 80% serving margin at THEIR prices lololololol
🚀 Introducing DeepInfer, our in-house inference framework, now fully open source!
Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. We’re sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable.
🔗 github.com/deepseek-ai/DeepI…
1/4
prompt_hater retweeted
New on the blog:
— Room for ambition
A late-night conversation at a startup office in Finland has stayed with me for almost twenty years.
Some thoughts on ambition, getting good at something, and the people around you.
karrisaarinen.com/posts/room…
prompt_hater retweeted
Skill: april-zhu.com/colorfigures
Demo: april-zhu.com/colorfigures-d…
Demo currently has 26 figures and I am adding more! Have fun
prompt_hater retweeted
I often get emails from junior engineers asking me what to do.
Here's what I just wrote to one of them
prompt_hater retweeted
a few of my predictions for the next 6 months of ai data:
- a monopolistic human data buy side layer emerges. right now you can use zap mail, heyreach, instantly and claude max plan to reach out to practically every potential human data buyer on the world in 7 days and (even with the worst strategy) at least come out with roughly 20 buyers spanning rl builders, neo labs and (if lucky) one or two major labs, who you could easily offload $20-$50m cumulatively from the very start. it is not going to be this easy to find buyers forever. as labs inevitably get f*cked by bad apples on rights, de-id, QC and the rest, a select few will emerge as “transaction gates” in which every corpora must flow them in order for the labs to deem as safe to buy. analogous to how a few labs already don’t touch corpora if it was de-id by Tonic Textual.
- the waterline continues to rise. ai labs are buying real company data to build artificial workspaces to teach their models how to do an office job. using a headcount as a metric, 3 months ago they bought company internals with 10+ headcount, now they almost exclusively purchase from companies with 50 or more employees. as models saturate these smaller workspaces, demand for smaller company corpora plummets whilst demand for the already dwindling supply of large sprawling company internals increases. yet doesn’t skyrocket as at a certain point if the price becomes too high then labs fall back on synthetic expert work. it goes without saying that small companies in a esoteric domain models underperform in that hold vertical depth (e.g a 30-man freight brokerage) will still hold immense value.
- reality gets graded more and more. as the most knowledgeable domain experts begin to fail in writing valid rubrics that aren’t instantly saturated we are going to have to get reality to play the verifier, domain expert and environment, if we want progress to continue. rubrics shift from the attributes of a perfect NDA to a lab measured critical tempurature.
- atom environments. why hasn’t the vertiginous increase of ai capabilities led to huge economic growth? how can we get non-tech people to understand how intelligent ai is already? how do we find the best value/token ratio? to “solve” these questions I believe need to give ai a “body” in which it can interact with the real world at extremely low bandwidth. giving ai a body sounds like a humanoid which is obviously just one small slice of real world interaction, a little bit like how literally none of the ways we interact with in the real world is only via our real body but via tools instead like handles, vehicles, instruments and cutlery. maybe I should reframe it from giving ai a body to giving the world ai
- protocols for company data emerge. for such a fuzzy yet crucial area of ai, it is insane how unorganized certain steps in company data brokerage are, such as giving context on what your companies internals actually hold.
prompt_hater retweeted
hard disagree.
there's a system in place whose sole purpose is to address these issues. as citizens, yes, it's our responsibility to raise our voices, contact whoever we can, and push for things to get fixed. but to what extent?
we all have jobs, responsibilities, and lives of our own. we can't just put everything aside to do the work of people who are literally responsible for handling these problems. it's the job of higher authorities to identify these issues and ensure they're resolved.
and idk if you realise this, but the only reason they're fixing it now is because a foreigner with public visibility has started raising questions. they know addressing it will earn them some good PR.
but why does something as basic as this even need to go viral to get fixed? there are official grievance portals, there are social media platforms, and people have been lodging complaints day after day, only to get absolutely nothing in return.
if a system only starts functioning when there's public outrage or an opportunity for good PR, then what's even the point of having that system in the first place?
makes me wonder who's the real parasite here.
I think what is actually needed is for Indian people to actually give a damn about their cities and fellow citizens.
Anybody using the "I pay taxes, I shouldn't have to do anything" excuse should go live in another country. India isn't developed, it's developing, so it needs patriots, not parasites. If this triggers you, it means you're part of the problem.
If you're not happy with the footpath (or a pothole, or anything else which is broken), fix it, or call someone to fix it, and don't give up until it's fixed. Be relentless. Be vocal. Be creative. Make it your mission to get that one thing fixed. Look at what Virendra (@Virendra0698) did by just calling the right people and being respectful (but also persistent). He and Arun Pai (@arunpaiblrwalks) along with the help of GBA and some wonderful volunteers turned an unofficial garbage dump into a pristine, beautiful footpath through sheer force of will.
xitter.cf/Virendra0698/status/21…
prompt_hater retweeted
The evergreen resource - An Introduction to Distributed Systems by Kyle Kingsbury
github.com/aphyr/distsys-cla…
For an impromptu discussion, the Google File System (GFS) came into the picture. And from there, I remembered this MIT course 6.824 Distributed Systems, where there is a lecture on it.
Many folks have covered GFS in depth, but this course also provides an in-depth view of Distributed Systems in general. Recommend this if you are interested in this domain.
nil.csail.mit.edu/6.824/2021…
(Note: the site is HTTP though)
prompt_hater retweeted
Give your agent a way to verify its work. This can dramatically improve the quality of output.
With this process in place, Claude Code was able to write 90%+ of its code itself. When generating software at a faster rate than you can review it, the bottleneck shifts to proving that it works. @bcherny and team invested enough in verification, we've been able to move fast while maintaining stability.
Code reviews aren't going away, but you need to think of it differently. One human reading through line-by-line is no longer going to work as a gate, it needs to evolve into a multi-layered verification system, built with fast deterministic feedback loops for your agents.
This is where e2e tests, feature maps (as @poteto has been advocating), property tests and multi-agent adversarial reviews come in. These all provide opportunities to catch errors early and fix them before things get out of hand.
You don't have enough eyeballs for the amount of code being written so your focus needs to be on product intent, risk assessment and evaluating evidence rather than reviewing code. Delegate the code checking work to your automated tests, scale your approvals based on risk, and turn your test failures into living evaluations. This approach helps maintain speed without losing reliability.
Let the agents do what they do best and let you focus on what makes you irreplaceable: judgement, direction, and ownership. We tried to capture how we handle code review and verification in this writeup. Hopefully it's helpful.
Claude writes about 90% of Claude Code's code.
From June to September, merged PRs more than doubled while the share reverted fell by more than 40%.
We wrote about how we changed code review to get there, and how to adopt it:
claude.dev/blog/solving-code…
prompt_hater retweeted
When I was writing Unreal Engine 1, my time was divided as follows:
- Around 30% went to seriously deep coding on hard problems: assembly language texture mappers, C++ BSP trees, data persistence, object oriented systems design, and networking.
- Around 70% was just typing and refactoring large bodies of code needed for a complete engine feature set. This was not hard, and while I was doing it my mind was mostly on the deeper coding topics, thinking and planning for what's next. It didn't feel much different from vibe coding today, except that it was just me typing code on an IBM Model M keyboard rather than prompting it and reviewing the results.
So, it has always seemed perfectly logical that we would rely on AI for that 70%, once the AI could do it with a similar quality standard as we could, which has never been perfection. It came as no surprise that, after a few years of AI model evolution, it was able to take the reigns on that 70% and free up programmers' time for more valuable tasks.
What's more of a surprise is the emergence of AI as a sort of insta-Carmack able to solve hard technical problems quite cleverly and often cleanly, and that it's able to reason about and work in complex 3D scenes both by operating the tools and by directly understanding the raw data files.
I abandoned assembly language in 1998 when out-of-order CPUs made much simpler C/C++ code faster, and never I missed it. If prompting through architectural documents, concept art, and design plans can transcend typing code and tweaking vertices, that will enable developers to spend their time on higher-level, higher value technical and creative tasks.
What matters is the end product, whether it's a game or an engine. Anything developers can do to try out more ideas and make better software and games faster is a win for everyone.
prompt_hater retweeted
【速報】pstack 本家がClaude Codeでも使えるようになった！！！！！！！！！
正真正銘、最高で最強のプラグインになったわけだ！！！！！！！！！
github.com/cursor/plugins/tr…
prompt_hater retweeted
Here's my take on what a software factory really is
This is my talk from @aiDotEngineer NYC, and I decided to record a version for everyone to watch on X.
tl;dr - software engineering is becoming factory engineering, and there's new work (especially factory benchmarking) that engineers should start focusing on.