Renat Lotfullin

Sustainable world evangelist |

[Renat Lotfullin @renatlb]

@User10371721742 @ProfBuehlerMIT IMO scientific breakthroughs are emerged from paradigm shift where not just modified/adjusted metrics rules applied but total concept transformation and thinking from completely different perspective

[Renat Lotfullin @renatlb]

@xbresson @NeurIPSConf Yes it is! Future of engineering and science AI are graph based

[Renat Lotfullin @renatlb]

@daleverett and the rust for future science graph-based AI 🔥

[Renat Lotfullin @renatlb]

@John_W_Maki @ProfBuehlerMIT Thanks a lot for brilliant ideas!
"Verifier gates" - that's exactly how it should be named

[Renat Lotfullin @renatlb]

As a KAN/GNN evangelist I was so fascinated listening to prof @ProfBuehlerMIT at his "Superintelligence for scientific discovery" webinar this winter.
An the newly published paper "Self-Revising Discovery Systems for Science..." is one more cornerbrick in science AI foundation

[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

[Renat Lotfullin @renatlb]

@ProfBuehlerMIT This paper is really masterpiece giving me answers for many science AI mechanics:
1) How we can mathematically evaluate invention novelty -> left KAN extension;
2) Artefacts typing -> typed provenance

[Renat Lotfullin @renatlb]

@ChiefScientist @nikalaikina а еще лучше Rust для графовых AI!

[dale @daleverett]

Even the most powerful artificial minds need infrastructure to understand the world.

We've launched a radical project to change the way we build AI Agents. A graph layer that uses up to 34x less RAM than neo4j.

To start: pgGraph (Full Open Souce, Rust),
is a postgres extension that lets you run complex graph queries, like finding shortest paths using standard SQL. Directly in postgres.

Coming next: Polygres,
give your postgres a warp drive by combining pgGraph & pgVector into an all-in-one database.

We're building the foundation of intelligence @evokoa_ai

Links below ↓

[Renat Lotfullin @renatlb]

@NateHindman @HilaShmuel Cabinet for collaborative work?

[Renat Lotfullin @renatlb]

@ATinyGreenCell I develop biobased materials and agriculture residuals bio processing

[Renat Lotfullin @renatlb]

@daleverett thx for sharing

[Renat Lotfullin @renatlb]

@BlancheMinerva @storytracer @soldni @jakepoznanski @status_effects @leppert @JJitsev Thanks a lot! That's so helpful I was struggling with old technical books parsing

[Renat Lotfullin @renatlb]

@HilaShmuel windows desktop 🙏🙏🙏🙏🙏🙏🙏

[Renat Lotfullin @renatlb]

@IntuitMachine

[Renat Lotfullin @renatlb]

@OneOrigine might be helpful with ui/ux, can't dm

[Renat Lotfullin @renatlb]

@AlexanderKalian @0x506c61746f so interesting, I'm experimenting context map approach over graphs

[Renat Lotfullin @renatlb]

@AlexanderKalian https://t.co/h1lVIG5b7M

[Louis Maddox @permutans]

“Higher-Order Graph Attention via Diversity-Aware k-Hop Sampling” (2026) https://t.co/jeakLk2vXf
HoGA https://t.co/v16g6hhSfH

> at least a 5% accuracy gain on all benchmark node classification datasets and outperforms recent baselines on 6 of 8 datasets.

[Renat Lotfullin @renatlb]

@ben_burtenshaw HF agents course any updates in future?

[Renat Lotfullin @renatlb]

@w2sgarnav interested in ps3, how to join?