👨💻 Senior AI Research Engineer @NVIDIA. Ex-Groq. Occasional angel investor. Opinions are my own.
Joined July 2009
- Tweets4.5K
- Following429
- Followers6.3K
- Likes10.7K
Rick Lamers retweeted
Wow that escalated quickly. Was gone for a few weeks and now pretty much everyone launched their own decision model copycats:
- Sep 23: Together AI, Tev1-4B
- Sep 24: Alibaba Cloud, decision-model
- Sep 29: Liquid AI, d1
- Sep 30: Databricks, ai_decide
- Oct 1: Amazon AWS, Strands Decider 2B
- Oct 1: Cloudflare, Clef and Clef-flash (Clef-omni on Oct 9)
- Oct 1: Perplexity, Decider (v1.3 on Nov 12)
- Oct 6: OpenAI, Decisions API (GPT-6 Luna)
- Oct 7: H2O.ai, H2O-Lightning-4B
- Oct 7: Snowflake, Snowflake Decision
- Oct 9: Microsoft, Microsoft-Decision-1
- Oct 15: IBM, Granite Switch 4.1 with Decision adapter
- Oct 21: Baseten, Mammoth-1 and Elephant-1
- Nov 6: Salesforce, Koa Decisions
- Nov 11: Google DeepMind, Gemini Decision 1
According to a Vercel report, 26% of requests are now routed through decision models 📈
From the reply:
> In case that wasn’t clear, the process is:
> - Generate an AI image of a living room
> - Add a real photo of a car in Photoshop
> - Display the altered image on a laptop
> - Take a real photo of the laptop
> - Upload that photo to Pangram
That's wild 🫡 @pangram
Rick Lamers retweeted
Each agent gets a skill library, motion planner, simulator & 10 demos. It writes skills, simulates ~30x, then asks for ~10 real trials
6 frontier models
5,400 real-world rollouts
Spatial generalization measured
Results: never-before-seen success rates and new behaviors! 2/12
Rick Lamers retweeted
Inkling 2 Small’s weights are now available for download on HuggingFace under an Apache 2.0 license.
Make Inkling your own!
huggingface.co/thinkingmachi…
a definition of AGI I hadn't seen yet (maybe a me issue) but enjoyed more than expected
> I believe we’re in the early stages of the singularity, defined as the period when the rate of technological and scientific progress exceeds the rate at which we humans can parse out and make sense of it
src: @jackclarkSF
1
264
JIT your Rust folks (esp during agentic dev!)
Rust has compile time issues. I have a fully optimized LLM accessible flow with Relax, a Rust JIT. Now with fable 5.2 I can finally have a worthy website for this absolutely jawdropping piece of technology made by machines, for machines. makepad.github.io/relax/
262
Rick Lamers retweeted
I had Codex build a CPython bytecode compiler in Rust (`serc`).
It emits byte-for-byte identical output on the top 10,000 PyPI packages, 2-4x faster than the CPython implementation.
This was entirely an experiment, though we may ship it (opt-in) in uv.
Rick Lamers retweeted
If your agent runs on an open-weight model, its reliability depends on a choice you may not have made deliberately: the inference provider.
We decided to make it visible 👇
“This model is bad at tool calling.”
Maybe. Or maybe it’s your provider.
Today we’re launching the Tool-call index, to help you find the best provider for each model 👇
providers.dottxt.ai
Rick Lamers retweeted
When Anthropic release a new model it's always worth taking a look at their updated prompting guide - here's an entertaining tip to make Fable 5.2 less likely to be cautious in how it improves existing code: tell it "This codebase is yours" platform.claude.com/docs/en/…
ALT Say who owns the code Claude Fable 5.2 responds closely to what the prompt says about who owns the code it works on. In Anthropic’s testing on small coding tasks, when a prompt said nothing about ownership or presented the code base as someone else’s, the model tended to add to the existing structure, even when that repeated existing code. It also tended to report nearby problems instead of fixing them. When the prompt says the code base is the model’s own, the model often merges repeated code and makes related fixes without being asked. If you want the model to improve the structure as it works, say that the code is its own: This codebase is yours: what it's for and how it's built are your call, and you'll keep building on it. In Anthropic’s testing on small coding tasks, the model merged repeated code in about half of the tries with this sentence, and in under a tenth of the tries without it. The rate varied by task. The result held when the sentence stood inside a long system prompt
No small feat! Congrats @RonanKMcGovern 👏
Our Dialt Fast s2s model is on the frontier!
Ahead of Gemini, OpenAI and ElevenLabs Live models!!!
W/ big thanks to @_josh_meyer_ for the independent eval from @veris_ai !
1
1
387
Rick Lamers retweeted
OpenCode 2 far outscores OpenCode 1.19: from 8th of 9 harnesses to tied for first on all three models.
It replaces 1.19 in the @huggingface Harness Leaderboard, which compares coding agents on the same models. Like most harnesses there, it now uses the Chat Completions API directly, so it's more comparable.
At high effort it solves 65.7% of tasks instead of 43.3% on GLM-5.3-Flash and 69.4% instead of 57.0% on DeepSeek-V4.1-Flash. On Flint, 22.9% instead of 7.1%.
Cost per solved task fell from $3.35 to $0.45.
Part of the old gap was the setup. Harbor's standard route for OpenCode 1.19 is OpenAI's Responses API, which I translated to Chat Completions so it works with Hugging Face Inference Providers.
It ran without errors but silently dropped the model's earlier reasoning. Thanks to @thdxr and @rekram11 from @opencode for spotting it!
Leaderboard: huggingface.co/spaces/joelni…
It’s nice that I can now explain @thinkymachines to my friends as the post training API company that absolutely slaps at pretraining (too)
Excited to support this incredible team as a small check investor 💪
Darwin Robotics Raises $3.6M to Build General-Purpose Robot Technicians
We're excited to announce that @darwin_robotics has raised $3.6M in pre-seed funding, led by @Link_Ventures, with participation from @ycombinator, @Liquid2V, @RebelFund_VC, and the co-founders of Open AI, xAI, and Y Combinator.
We're building general purpose robot technicians that assemble, commission, and maintain critical infrastructure. Our initial focus is AI infrastructure, where rapid data center buildout is creating growing demand for skilled physical work.
Every leap in human progress has been driven by our ability to build increasingly sophisticated machines. The next frontier is intelligent robots that can use tools to build and maintain other machines.
Our robots learn complex tasks from human demonstrations rather than being explicitly programmed for each workflow. We're deployment-first, working closely with AI infrastructure partners to bring our first robot technicians into real-world environments.
Our mission: Robots to build civilization.
We're hiring exceptional engineers and researchers in robot learning and hardware, and seeking deployment partners building and operating data centers.
darwinrobotics.ai
1
3
384
Rick Lamers retweeted
Z.ai chief scientist Jie Tang just revealed the training pipeline behind the upcoming GLM-5.5 (32B, 122B, and 256B) during a lecture at Tsinghua University today.
Instead of relying on public datasets, their pretraining pipeline pulls from six separate internal engines:
- Web: 1 trillion pages crawled at ~10B pages/day with PB-scale template clustering
- Documents: 2 billion scientific PDFs parsed through GLM-OCR (4x faster with speculative decoding)
- Books: 50 million books with layout reordering and garbled-text filtering
- Code: 500 million repositories turned into executable tasks with real tests and issue descriptions
- Video: 10 billion videos for long-sequence world knowledge
- Audio: 300 million hours rewritten via ASR into trillions of tokens
Awesome harness eval, TLDR pi sits at Pareto optimal: not highest accuracy, but lowest cost per solve at quite acceptable accuracy
mini-swe-agent the paper-cited capability measurement harness gets the highest accuracy at a pretty nontrivial cost increase (which I think fits its target use as an eval harness)
Claude Code and Codex are middle of the road with CC beating BOTH accuracy & cost per task of Codex
I burned a shitload of compute so you don't have to. Releasing the Harness Leaderboard.
Which harness should you run your open model in? I ran 28,849 trials to find out.
The leaderboard compares nine coding-agent harnesses on open models. In each comparison only the harness changes: the model, provider and reasoning effort stay fixed, and every task gets three attempts.
The trials ran GLM-5.3-Flash, DeepSeek-V4.1-Flash and Flint on Terminal-Bench 2.1 and DeepSWE, for $4,746 of inference.
Averaged over the three models, mini-swe-agent (@KLieret), a bash-only agent, solves 51.4% of tasks and @openclaw 27.3%. Four others are statistically tied with mini-swe-agent.
One of them is Pi (@pidotdev), at 49.4% for $0.29 per solved task, about 45% less than mini-swe-agent's $0.53.
The best harness depends on the model: mini-swe-agent on GLM-5.3-Flash, Hermes Agent (@NousResearch) on DeepSeek-V4.1-Flash and Kimi Code (@Kimi_Moonshot) on Flint.
@opencode costs $3.35 per solved task, 11.7 times as much as Pi.
See Harness Hub's Harness Lens for what the nine harnesses do differently. It documents the prompts, tools and compaction of 34 coding-agent harnesses and links every claim to its source.
Leaderboard: huggingface.co/spaces/joelni…
Harness Lens: huggingface.co/spaces/harnes…
9
8
634
Superexponential fit fits better than the exponential fit 🤔
3. Interviews and surveys with staff are consistent with AI usage driving current acceleration, though this evidence is very hard to interpret. Staff self-report dramatic productivity gains, but say the most important AI R&D contributions aren’t yet fully automated.
1
1
408
Rick Lamers retweeted
We've been working towards AI image and video detection for months, and now it's finally here.
There were a lot of challenges along the way. There are way more AI image/video models than frontier LLMs, and their outputs are more diverse. Most phone cameras, including iPhones, do some level of AI postprocessing to add detail to real photos.
We had to build a model that's robust to this postprocessing while understanding the difference between AI edits and fully AI-generated visuals.
I think the heatmap feature is a really important step forward and a huge differentiator over other models which typically report scores as a binary confidence.
Please try it out!!
