
Autonomous Research Agents | Founder of Agentic AI Platform @Uber | Ex-Founder @onepanelai, Kubernetes-native AI Platform
United States
Joined September 2011
- Tweets658
- Following77
- Followers156
- Likes500
Rush Tehrani retweeted
LLM next-token prediction scales because text comes from a discrete, fixed vocabulary.
Audio doesn't - it's a continuous waveform signal. @liquidai LiquidCodec is our new tokenizer for sound, unlocking audio generation directly within discrete autoregressive language models.
Convert continuous audio into a discrete vocabulary, train on the tokens, and decode back the raw audio.
waveform → tokens → waveform 🚀
Introducing 𝙻𝚒𝚚𝚞𝚒𝚍𝙲𝚘𝚍𝚎𝚌, our ~74M-parameter streaming audio codec powering LFM3-Audio-2B and LFM3-Omni-2B.
An audio codec turns speech into tokens a language model can predict, then converts those tokens back into audio. We built ours around two requirements: tokens that are easy to predict and fast conversion back into speech.
On a single CPU thread on an Apple M4 Max, LiquidCodec runs at approximately:
> 361× real time for encoding
> 87× real time for decoding
Both measurements use FP32 precision.
1/ 🧵
Very solid guide on when to use Anthropic’s models
Which Claude model should you build with?
Our new guide breaks down when to use Fable 5.2, Opus 5.5, Sonnet or Haiku. It also covers when to adjust effort instead of switching models, and how to save cost by putting subagents on smaller models.
claude.dev/blog/which-claude…
1
27
Big improvements in robotics control with Fable 5.2
Replying to @claudeai
Fable 5.2 shows substantially improved skills in controlling physical robots, even on tasks it has never explicitly been trained to do.
On RoboDojo, a benchmark with a simulated two-arm robot, it outperformed other models and most AI systems built to control robots.
67
Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges.
They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.
Rush Tehrani retweeted
Memory is what allows Claude to learn from experience. In a new Engineering Blog post, we share how “dreaming” helps Claude organize what it’s learned, correct its mistakes, and find what it needs more easily the next time. anthropic.com/research/compo…
I'm delighted to partner with @LaudeInstitute (along with my @Sanjay_Ghemawat, John Hennessy, Dave Patterson, Ben Gomes and @andykonwinski) to support funding of open science moonshots!
anyone who is close to research knows how broken the incentives are. doing open science at the frontier is harder than ever, compute costs alone price most teams out. we kept wondering what might happen if we give the right resources to the right researchers at the right time?
laude.org/updates/moonshots-…
Rush Tehrani retweeted
Real time Atlas! Try it now! 🚀
Rush Tehrani retweeted
We’re taking quantum communications to new heights! ✈️
Late last month, @NASA researchers made history when they successfully sent tiny light particles, or photons, between an aircraft and a ground station over a quantum channel, the first time this has been achieved in the U.S. (1/3)
Rush Tehrani retweeted
We’re sharing how our biology research team worked with Claude to map the first 3D structure of a vault particle from bacterial life.
Vault particles are a longstanding mystery. They are barrel-shaped objects, among the largest protein structures found in our cells. They were first discovered by Leonard Rome and Nancy Kedersha in 1986, but their function remains unknown.
With Claude, we’ve found results that might bring us closer to solving that mystery.
Until now, vaults were only known in the cells of complex things like animals and plants. We have shown that similar proteins from bacteria (and the viruses that infect them) can also assemble into vaults. Finding the structures in these simpler forms of life may make it easier to investigate their function in our own cells.
For our life sciences team, the project was also a demonstration of the potential (and the limitations) of using AI to accelerate science.
In our full piece, we discuss how Claude was able to dramatically speed up some stages—manually determining the vault particle’s structure would have taken weeks to months of painstaking work, but here it took days. We also discuss how Claude made some simple mistakes along the way, and how we collaborated with our machine learning researchers to address them.
Read more: anthropic.com/research/inves…
Rush Tehrani retweeted
There's a buzzy new paper about GPT-6 Astra being able to directly interpret model weights (prompts full of floating point numbers) - I tried it with Qwen 3.8 27B on my laptop and found that, after 45+ minutes, that model could achieve the same thing simonwillison.net/2026/Nov/3…
Rush Tehrani retweeted
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
Rush Tehrani retweeted
What if an optimizer could think & react while a neural network trains?
MIT researchers developed OPAL: a deep learning optimizer powered by an AI agent that watches the training dynamics & rewrites its own source code as training goes. It improves model performance across vision & language tasks, invents specialized schedules, & develops strategies to prevent loss spikes & escape overfitting: tinyurl.com/yfkekat2
Try it out for yourself: tinyurl.com/mttj5exh
With coding agents, much of the research and reasoning behind design decisions never makes it into specs or code. It becomes tribal knowledge between developers and their agents.
I’ve started committing decision logs to a shared repo.
Are others doing this? Should we?
5
1
107
Rush Tehrani retweeted
1/4
Meet Qwenputer: Qwen3.5-4B with a full RISC-V CPU inside its weights, using the model's own Gated DeltaNet layers as RAM.
It runs C and MicroPython in its forward passes, no external runtime.
On DeepMind math: base 55–90%, base + Python tool 95–100%, Qwenputer 97.5–100%.
TPU in orbit just wrote a haiku and sent it back. Wild.
A ‘hello world’ milestone for Project Suncatcher 🚀
After launching our prototype satellite a few weeks ago, we’ve now successfully sent a command up to orbit, booted up the TPU, ran the prompt, and received our first complete response back on earth - a haiku!
1
72























