Anirudh Ravichandran

MTS @ PrimeIntellect

[Anirudh Ravichandran @anravich94]

@willcb can attest - setup looks neat and silent enough that I can't hear sitting in the adjacent desk when @willcb does his thing and whispers hissing parseltongue with codex.

[Prime Intellect @PrimeIntellect]

Join us at Prime Intellect to build open superintelligence and the infrastructure powering self-improving agents.

We’re hiring across 25+ roles.

Research
• AI Research Resident
• Research Engineer, Distributed Training
• Research Engineer, Reinforcement Learning
• Research Engineer, RL Infrastructure

Compute
• Compute Finance & Strategy
• Compute Intelligence Engineer
• Head of Compute
• Solutions Architect, AI Infrastructure
• Technical Account Manager, AI Infrastructure

Engineering
• MTS, Compute Platform
• MTS, Full-Stack
• MTS, GPU Infrastructure
• MTS, Inference
• MTS, Sandbox Platform
• MTS, Security
• MTS, Training Platform

Growth
• Applied AI: Product Strategy & Revenue
• Forward-Deployed AI Strategy
• Head of Growth
• Head of Marketing
• Community and Devrel

Other
• Internship
• Open Application for Unconventional Talent

Apply here: https://t.co/eIImvMMnqb

[Anirudh Ravichandran @anravich94]

Every few days, @mikasenghaas cooks up something better than the previous week. Amazing work.

[Mika Senghaas @mikasenghaas]

we just shipped adaptive concurrency to prl main, making it (to the best of my knowledge) the first oss rl framework to dynamically adjust inflight rollouts to maximize throughput during the entirety of an rl run

[Anirudh Ravichandran @anravich94]

One huge step towards RSI.
@eliebakouch cooked 👨‍🍳

[Prime Intellect @PrimeIntellect]

We ran the largest open experiment on how frontier models do AI research.

100+ autonomous runs across 10+ models, sandboxed on 8xH200s for up to 8 days, iterating on the nanoGPT optimizer track.

Best runs closed 82% of the gap to a record built by dozens of humans over months.

[Anirudh Ravichandran @anravich94]

Good read

[Abhay Singhal @_AbhaySinghal]

https://t.co/bs3lwEpQD9

[Aravind Srinivas @AravSrinivas]

Impressive numbers for a 700b parameter model!

[samsja @samsja19]

It was a pleasure to collaborate with the talented @mixedbreadai team to push a sota agentic search with our prime rl stack

RL is truly beautiful.

Even the biggest model don't adapt well to new harness and new tool out of the box, but with RL you can get the model to outperform extremely well at a faction of the cost

Everybody spending meaningfully amount of money on inference will eventually need to postrain

[will brown @willcb]

we're accelerating

[David Ondrej @DavidOndrej1]

how Prime Agent learns

compared to Hermes

[Anirudh Ravichandran @anravich94]

We're hiring btw. DM or apply: https://t.co/oHMkSfi6y7

[Prime Intellect @PrimeIntellect]

Introducing Prime Agent:

A self-improving RLM harness for coding and long-running autonomous tasks.

Designed to be both token-efficient and expressive through programmatic tool calling, context as a variable, multi-agent messaging, and a self-modifiable harness state.

[Daniel Auras @rasdani_]

btw there's only one team i trust to train on this beast of a harness

[Anirudh Ravichandran @anravich94]

Checkout this super cool work

[Prime Intellect @PrimeIntellect]

Introducing Prime Agent:

A self-improving RLM harness for coding and long-running autonomous tasks.

Designed to be both token-efficient and expressive through programmatic tool calling, context as a variable, multi-agent messaging, and a self-modifiable harness state.

[Aravind Srinivas @AravSrinivas]

Two orders of magnitude improvements are quite rare. This is a big deal.

[Anirudh Ravichandran @anravich94]

I agree, though I think methodically done synthetic data with light human touch does scale significantly better. Ironically, you'll need post-training researchers to do that.

[altra @catboosted]

Just a warning to any would-be founders trying to get into RL/data:

You’re signing up for pain. It’s not scalable. Or durable.

You’re trading man-hours for a 30-40% margin. Revenue only scales with human experts/eng talent.

Tell me the staff headcount, I’ll tell you the rev

[Anirudh Ravichandran @anravich94]

This is actually a model that punches above it's weight. It definitely seems like it's in a very nice spot on the pareto cost-perf curve.
On a lot of benchmarks that you'd generally care about, it does reasonably well and looks like a pretty adaptable model to post-train.

[Mira Murati @miramurati]

Inkling-Small is comparable to Inkling at a quarter the size. Weights are open, fine-tunable on Tinker today. Look forward to seeing what people make with it.

[Ishaan @auto_grad_]

@Shekswess @arcee_ai @awscloud @PrimeIntellect great work sir!

[Anirudh Ravichandran @anravich94]

Happy to see this! 🇮🇳

[Hemant Mohapatra @MohapatraHemant]

Announcing a San francisco office and great to bring in @dchaplot as a hands on advisor to help @SarvamAI build frontier models from India.