Alex Krentsel
@AlexKrentsel
Autonomous Systems PhD @Berkeley_EECS, Sr Researcher @GoogleResearch.
systems guy working on harness-layer self-improvement
251 Following 4.2K Followers
Agent harnesses are usually hand-tuned, or optimized offline against benchmarks. What if an agent could improve its own harness *while doing the work*?
Introducing Exo, a harness architecture for runtime self-improvement 🧵
(paper:
Show more
i've reproduced the results from the MIT blog post today on weight-space programming. but one additional result i have to add: this is a new, step-change emerging capability that's appeared in just the latest set of models.
Show more
I do not understand how there continue to be *more* companies popping up doing cloud sandboxes for agents. how does this happen? exponentially growing number of companies?
this is just going to be commoditized... someone will build a layer for this... (SkyBox?)
Show more
very excited to see this paper get recognized, I believe the technique here is widely applicable.
This was joint work done with the SkyDiscover team (@shagarw21, @mertcemri, @shulynnliu), along with our advisors (@istoica05, @matei_zaharia, Mohsen Lesani, and Sylvia Ratnasamy),
Show more
super excited that our paper on Inductive Deductive Synthesis just got accepted to NeurIPS as 🏅Oral🏅 (top 0.34%)
in the paper, we show how given a formal Lean/Rocq spec, we can incrementally co-synthesize *proof* with the *code*, beating direct code synth
link &takeaways⬇️🧵
Show more
Formal verification × agents 🚀
Our IDS paper was accepted as a NeurIPS oral! We study how agents can co-evolve code + proofs, achieving 3x the success rate of Claude Code on distributed systems specs.
IDS Paper:
Code:
And also check out our recent work on the next challenge: ensuring the spec itself captures human intent. 👇
Show more
I used Opus 5.5 to formally verify the Claude Agent SDK using Lean. A couple short prompts = 16 PRs fixing various bugs and race conditions. Video attached.
TLA+ also works well. I sometimes combine Lean and TLA+ to look for issues around data flow, concurrency, and state mgmt.
I don't know either language well, but Claude is excellent at both. This approach is super useful for formally modeling your code and finding bugs that a human probably wouldn't have spotted.
Is formal verification the future of coding (or at least, bug finding)?
Show more
super cool work from Mayank. excellent blog post on how to reach very impressive performance with orders of magnitude less compute
We pretrained a 2.3B MoE (360M active) Hybrid Mamba-2 that lands within a few points of Llama-3.2-3B using <1% of its pretraining FLOPs.
No dedicated cluster. The run hopped between H100s, A100s, V100s (yes, V100s) and TPU v5p/v6e on a single codebase.
Meet Rigel 🧵
Show more
Introducing Exo 🦋 – an open-source agent that can rewrite every part of itself. It learns new environments, adapts policies, even optimize costs (and play games!!)
We've been hacking away at this project with @martin_casado from @a16z and @ankrgyl from @braintrust and friends.
Show more
