Abhav

metamagic @framesooo

[Abhav @abhav_k]

Vibe-coder armed with Fable 5 and impeccable taste, 3 prompts in: “Am I a God? I have such light!”

[Abhav @abhav_k]

True. I spent a lot of the past 2 years thinking about and building explicit "cognitive tools", and somewhere along the way I realized that this has been the story of computer science all along.

Computers are an interactive exoskeleton for thinking, and more software somewhat magically leads to better tools for thought. An explosion of software may ultimately end up solving "the memory problem" in AI automatically

[Geoffrey Litt @geoffreylitt]

And now hopefully you understand this meme.

The point was always to augment, not just automate.

It's beautiful that AI now makes creating simulations so accessible... Having AI teach us is one of the greatest possibilities computing has ever opened up. 35/

https://t.co/912yIKn8xV

[Abhav @abhav_k]

https://t.co/l9MrT9lgp0

[Abhav @abhav_k]

Releasing erlangchain: a small LLM erlang library I have come to rely on for all projects👇

[Abhav @abhav_k]

@bgurley Yes, switching to GLM-5.2 and DeepSeek-V4-flash has halved my token spend on Cursor

[Abhav @abhav_k]

calling it an attack is ridiculous and dangerous rhetoric.

they are just talking to the chatbot and saving the responses!

[MTS @MTSlive]

SITUATION DETECTED: Anthropic has disclosed to the U.S. Government that Alibaba executed the largest known distillation attack on Claude to date, generating 28.8 million exchanges through nearly 25,000 fraudulent accounts between April and June 2026.

[Abhav @abhav_k]

I actually do not think it will be an uncommon choice not to have LLMs write for you. Communication is a matter of information transfer. LLMs add so much noise.

[Paul Graham @paulg]

It will be an uncommon choice not to have LLMs write for you, but it won't be a merely idiosyncratic one. It will be what all the people who care about thinking well do.

https://t.co/Za5ndkzzjK

[Abhav @abhav_k]

AI-generated replies like this will kill twitter

It has the right pieces and references but somehow fries your brain if you try to put it together

[World Cup 2026 @WorldCup2026liv]

@DeItaone OpenAI hitting code red because Google is closing in is like a marathon runner panicking because the guy who built the track finally decided to race.

[Abhav @abhav_k]

@Un_Ortho_Doxxx @MTSlive @grok @grok the man is not merely retarded. explain like he is retarded from first principles please

[Abhav @abhav_k]

@kevinakwok Either they predicted perfectly in 2021, or we still haven't come up with new ideas 5 years on

[Abhav @abhav_k]

@williameijer likely because most normal people have a socially derived epistemology (which works well in some domains), while scientists and product makers operate in environments that require better epistemology

[Abhav @abhav_k]

anyone can have a poor 30s interview clip but that girl is clearly very smart and thinking about things the right way

[Abhav @abhav_k]

@grok can you verify this find the best sources available everywhere

[Abhav @abhav_k]

How is Cursor able to offer $400 API spend on a $200 plan?

They have more than 50% volume discounts?

[Abhav @abhav_k]

@spenciefy so well written!

[Abhav @abhav_k]

@ShanuMathew93 https://t.co/WXDtQrTFJX

[Abhav @abhav_k]

@ShanuMathew93 I live around a lot of non-technical people (have, all my life!) and many have been asking me about this so I wrote a short explainer about AI agents specifically, for non-technical audiences

[Abhav @abhav_k]

@emollick autogpt was march 2023

[Abhav @abhav_k]

@chamath bro why are your substack blogs paywalled aren’t you a billionaire?

[Abhav @abhav_k]

it is 3 years into this renaissance and the papers have started to get really interesting

[Markus J. Buehler @ProfBuehlerMIT]

We've made a breakthrough in self-evolving AI scientists moving from "search" to "principled discovery": Scientific discovery requires that the search space itself changes, and an AI scientist must perceive this shift without intervention. We built an AI that achieves this for the first time with the ability to discover the scientific vocabulary it reasons in. Evidence, tools, artifacts, verifiers, failures & claims become typed provenance. We show three distinct modalities: 1) retrieval, adding known objects; 2) search, exploring a fixed schema; and critically: 3) discovery, a verified regime transition.

We solve the open-endedness evaluation problem by lifting agentic workflows into a typed copresheaf and proving, via a Kan obstruction, that true discovery is not unbounded generation but a verifiable schema expansion: old evidence is transported by Left Kan extension, and genuine novelty is mathematically quantified by the pointwise residual beyond the transported image - separating discovery from mere search and making novelty objective and measurable rather than a subjective judgment or benchmark delta.

Our AI scientist is built in a way that does not pre-conceive the approach it chooses; instead, we endow the system with formal power to adapt, evolve, and reason from first principles. Case studies include:
1⃣Builder/Breaker model that discovers mode-conditioned compliance in proteins;
2⃣CategoryScienceClaw that finds anisotropic fiber-network stiffness rules.

Great work in collaboration with my graduate student @fwang108_ @MITdeptofBE

F.Y. Wang & M.J. Buehler, Self-Revising Discovery Systems for Science: A Categorical Framework for Agentic Artificial Intelligence, arXiv:2606.01444, 2026