Jit Ray Chowdhury retweeted
4 years ago, Zeromatter started with a single bet: building autonomous machines would be faster, safer, and better done first in a virtual world.
Today, we’re coming out of stealth to show what that bet has become.
Jit Ray Chowdhury retweeted
🚀 Autoware 2.0 is here!
End-to-end AI, hybrid driving stacks and new tools to discover packages, manage models and compose autonomous driving systems.
🔗Explore the release:
bit.ly/4BFynCp
#Autoware #OpenSource #AutonomousDriving
Jit Ray Chowdhury retweeted
The Era of Evals is coming to AI robotics.
Today we’re launching Reality Check: our leaderboard & first public robot manipulation benchmark.
14,400 real-world rollouts. Four models. Multiple tasks & data regimes. Object placements. Confidence intervals. FR3 Duo stations.
1/10
Jit Ray Chowdhury retweeted
Physical AI doesn’t just need more data. It needs refined data.
Validated. Calibrated. Traceable. Built with provenance.
That’s why we built Deepen Refinery: a marketplace for real-world data you can actually trust.
refinery.deepen.ai/
Jit Ray Chowdhury retweeted
New art project.
Train and inference GPT in 243 lines of pure, dependency-free Python. This is the *full* algorithmic content of what is needed. Everything else is just for efficiency. I cannot simplify this any further.
gist.github.com/karpathy/862…
Jit Ray Chowdhury retweeted
Since xAI was formed just 30 months ago, the small and talented team has made remarkable progress.
The future has never looked more exciting!
Jit Ray Chowdhury retweeted
The @xAI team is making progress
One year ago, Grok was only available on 𝕏, with no website and limited access
Fast forward to today, and Grok has expanded across platforms with rapid feature rollouts and major upgrades
New models. New apps. New features. Big upgrades Constant shipping
This kind of progress does not happen without an insanely driven team
xAI is moving at an unmatched pace, shipping updates faster than almost anyone in AI right now
Jit Ray Chowdhury retweeted
Grok
xAI’s new Grok Voice Agent is the new leading Speech to Speech model, surpassing Gemini 2.5 Flash Native Audio and GPT Realtime in our Big Bench Audio benchmark
The new model achieves a score of 92.3% on Big Bench Audio, just ahead of the previous leader, Google’s Gemini 2.5 Flash Native Audio Thinking. This model is @xAI’s first public Speech to Speech API, bringing increased competition to the space. The model has tool calling support and xAI has said it’s ready to be used across voice assistants, phone agents, and interactive voice applications.
Benchmark context: Big Bench Audio is the first dedicated dataset for evaluating reasoning performance of speech models. Big Bench Audio comprises 1,000 audio questions adapted from the Big Bench Hard text test set, chosen for its rigorous testing of advanced reasoning, translated into the audio domain.
Performance:
➤ Reasoning: Achieves 92.3% on Big Bench Audio, setting a new state-of-the-art for native Speech to Speech reasoning. Congratulations @xai and @elonmusk on this impressive release!
➤ Latency: At an average time to first token of 0.78 seconds, it is the third fastest model on our leaderboard behind Google’s Gemini 2.5 Flash Native Audio Dialog and Gemini 2.5 Flash Live
➤ Price: Simple pricing of 5 cents per minute connected, or $3 per hour of audio
Key features:
➤ Tool calling: Use built-in tools such as web search, RAG-powered search, or define your own tools with JSON schema
➤ Telephony: Connect to Session Initiation Protocol (SIP) providers like Twilio and Vonage
➤ Multilingual: Converse in over 100 languages with 5 voices to choose from
Jit Ray Chowdhury retweeted
nanoGPT - the first LLM to train and inference in space 🥹. It begins.
We have just used the @Nvidia H100 onboard Starcloud-1 to train the first LLM in space!
We trained the nano-GPT model from Andrej @Karpathy on the complete works of Shakespeare and successfully ran inference on it.
We have also run inference on a preloaded Gemma model, and we plan to try more exciting models in the future.
Getting the first H100 to work in space required a lot of innovation and hard work from the incredible Starcloud team to make this breakthrough.
This is a significant first step toward moving almost all computing off Earth to reduce the burden on our energy supplies and take advantage of abundant solar energy in space! 🚀
Jit Ray Chowdhury retweeted
Introducing Grok 4.1 Fast and the xAI Agent Tools API.
Grok 4.1 Fast is our best tool-calling model to date. With a 2M context window, it shines in real-world use cases like customer support and deep research.
x.ai/news/grok-4-1-fast
Jit Ray Chowdhury retweeted
We've raised $7M to help companies build AI agents that actually learn and work.
@Osmosis_AI is a platform for companies to fine-tune models that outperform foundation models with reinforcement learning.
Better, faster, and cheaper.
Jit Ray Chowdhury retweeted
Excited to release new repo: nanochat!
(it's among the most unhinged I've written).
Unlike my earlier similar repo nanoGPT which only covered pretraining, nanochat is a minimal, from scratch, full-stack training/inference pipeline of a simple ChatGPT clone in a single, dependency-minimal codebase. You boot up a cloud GPU box, run a single script and in as little as 4 hours later you can talk to your own LLM in a ChatGPT-like web UI.
It weighs ~8,000 lines of imo quite clean code to:
- Train the tokenizer using a new Rust implementation
- Pretrain a Transformer LLM on FineWeb, evaluate CORE score across a number of metrics
- Midtrain on user-assistant conversations from SmolTalk, multiple choice questions, tool use.
- SFT, evaluate the chat model on world knowledge multiple choice (ARC-E/C, MMLU), math (GSM8K), code (HumanEval)
- RL the model optionally on GSM8K with "GRPO"
- Efficient inference the model in an Engine with KV cache, simple prefill/decode, tool use (Python interpreter in a lightweight sandbox), talk to it over CLI or ChatGPT-like WebUI.
- Write a single markdown report card, summarizing and gamifying the whole thing.
Even for as low as ~$100 in cost (~4 hours on an 8XH100 node), you can train a little ChatGPT clone that you can kind of talk to, and which can write stories/poems, answer simple questions. About ~12 hours surpasses GPT-2 CORE metric. As you further scale up towards ~$1000 (~41.6 hours of training), it quickly becomes a lot more coherent and can solve simple math/code problems and take multiple choice tests. E.g. a depth 30 model trained for 24 hours (this is about equal to FLOPs of GPT-3 Small 125M and 1/1000th of GPT-3) gets into 40s on MMLU and 70s on ARC-Easy, 20s on GSM8K, etc.
My goal is to get the full "strong baseline" stack into one cohesive, minimal, readable, hackable, maximally forkable repo. nanochat will be the capstone project of LLM101n (which is still being developed). I think it also has potential to grow into a research harness, or a benchmark, similar to nanoGPT before it. It is by no means finished, tuned or optimized (actually I think there's likely quite a bit of low-hanging fruit), but I think it's at a place where the overall skeleton is ok enough that it can go up on GitHub where all the parts of it can be improved.
Link to repo and a detailed walkthrough of the nanochat speedrun is in the reply.
Jit Ray Chowdhury retweeted
AI & Robotics will have big impact fastest in dirty, dangerous, and dreary jobs
Dirty like garbage sorting
Dangerous like installing heavy solar panels
Dreary like butter passing jobs like ledger reconciliation or accounts receivable
First to cash flow positive wins
Jit Ray Chowdhury retweeted
I really wonder why so many AI companies seem to blindly compete for building the same thing: the best foundational models, as measured by some abstract set of benchmarks. It feels like there is now an entire ecosystems of companies that want to outperform GPT.x. Why?
Jit Ray Chowdhury retweeted
Waymo is now available to everyone in SF, no invite code needed. The future is here.
Now let’s make sure SF bureaucrats don’t ban it.
waymo.com/blog/2024/06/waymo…
Jit Ray Chowdhury retweeted
Had an insightful conversation with @geoffreyhinton about AI and catastrophic risks. Two thoughts we want to share:
(i) It's important that AI scientists reach consensus on risks-similar to climate scientists, who have rough consensus on climate change-to shape good policy.
(ii) Do AI models understand the world? We think they do. If we list out and develop a shared view on key technical questions like this, it will help move us toward consensus on risks.
I learned a lot speaking with Geoff. Let’s all of us in AI keep having conversations to learn from each other!

























