We launched a Claude Code learning mode! Claude Code not only makes you more productive, but can now also help you get better at coding. Whether you're a CS student or seasoned programmer, it will push you to think deeper about the code you're generating.

it's funny that chatgpt 5 will still struggle with such simple things like converting time.

super cool to compare the outputs from GPT-1 through GPT-5, given the same prompt: progress.openai.com

AI in HR: in an experiment with 70,000 applicants in the Philippines, an LLM voice recruiter beat humans in hiring customer service reps, with 12% more offers & 18% more starts. Also better matches (17% higher 1-month retention), less gender discrimination & equal satisfaction.

Sun's out, models out. 😎 IBM & NASA dropped Surya, an open-source heliophysics model trained on 14 years of observations from NASA’s Solar Dynamics Observatory, and it's 🔥🔥🔥.

Introducing ComputerRL, a framework for autonomous desktop intelligence that enables agents to operate complex digital workspaces skillfully. arxiv.org/abs/2508.14040 ComputerRL features the API-GUI paradigm, which unifies programmatic API calls and direct GUI interaction to

This is actually a really solid context engineering template. Kudos, Anthropic

GPT-5 just casually did new mathematics. Sebastien Bubeck gave it an open problem from convex optimization, something humans had only partially solved. GPT-5-Pro sat down, reasoned for 17 minutes, and produced a correct proof improving the known bound from 1/L all the way to

my experience with claude code frontend: quite awesome (no writing css on my own) backend: disaster - still have to read almost every bit of code, fix bugs

Paul Graham on why startups should deliberately choose hard problems:

One of the most prominent AI stars, who was among the handful who started the neural network revolution 15 years ago, and thanks to whom, we have all the wonderful models of today, is deeply sick. The person just disappeared, and no one among their colleagues spoke about them

planet earth is amazing US blocks china access to high end nvidia chips under constrains, china invents more efficient training techniques for LLMs china open sources everything a gain - for everyone in the planet.

OpenAI just claimed to have found the culprit behind AI hallucinations.

My house has 33 GPUs. > 21x RTX 3090s > 4x RTX 4090s > 4x RTX 5090s > 4x Tenstorrent Blackhole p150a Before AGI arrives: Acquire GPUs. Go into debt if you must. But whatever you do, secure the GPUs.

🎙️ Meet Qwen3-ASR — the all-in-one speech recognition model! ✅ High-accuracy EN/CN + 9 more languages: ar, de, en, es, fr, it, ja, ko, pt, ru, zh ✅ Auto language detection ✅ Songs? Raps? Voice with BGM? No problem. <8% WER ✅ Works in noise, low quality, far-field ✅ Custom

Handcrafted backend Vibecoded frontend

Technical founders should team up with cofounders 10x more technical and go after the hardest problems.

Updated & turned my Big LLM Architecture Comparison article into a narrated video lecture. The 11 LLM architectures covered in this video: 1. DeepSeek V3/R1 2. OLMo 2 3. Gemma 3 4. Mistral Small 3.1 5. Llama 4 6. Qwen3 7. SmolLM3 8. Kimi 2 9. GPT-OSS 10. Grok 2.5 11. GLM-4.5

A big part of our mission at Thinking Machines is to improve people’s scientific understanding of AI and work with the broader research community. Introducing Connectionism today to share some of our scientific insights.