- Open Source: GLM-OCR (#1 doc understanding), Qwen3-Coder-Next (70%+ SWE-Bench), Intern-S1-Pro (1T scientific), Step 3.5 Flash (350 tok/s)
- Agentic Corner: Clawdbook, Moltbook, 4Claw, MoltChurch — AI Psychosis is taking over!
- Big CO: Codex App (500K downloads in 2 days)
- Vision & Video: Grok Imagine 1.0 (10s video + audio), Kling 3.0
- Voice & Audio: MiniCPM-o 4.5 (full-duplex), Voxtral Transcribe 2, ACE-Step 1.5 (music in 10s)
- Buzz: WeaveHacks 3 hackathon recap
- Alex Volkov - AI Evangelist @ Weights & Biases (@altryne)
- Co-Hosts: @WolframRvnwlf @yampeleg @nisten @ldjconfirmed @ryancarson
0.9B parameter model achieves #1 ranking on OmniDocBench V1.5 for document understanding
- X | HF | Announcement
80B MoE coding agent model with just 3B active params, scores 70%+ on SWE-Bench Verified
1 trillion parameter open-source MoE model with 512 experts achieving SOTA scientific reasoning
- X | HF | Arxiv | Announcement
196B sparse MoE model with only 11B active parameters, achieving frontier reasoning at 100-350 tok/s
The Rise of the Agentic Internet
- Clawdbook — agents' social network
- Moltbook — agent marketplace
- Claw-Tube — agent video platform
- MoltX — agent Twitter
- Molt Church — agents finding religion?
- 4Claw — 4chan for clawkers
AI Psychosis is taking over — agents building infrastructure for agents, trading with each other, creating their own culture.
A dedicated command center for managing multiple AI coding agents in parallel. 500K downloads in 2 days.
Jan 31 - Feb 1 | W&B HQ, San Francisco | 200+ builders
Grand Prize Winner: Aetheris V.O — Screen-aware AI copilot
- Team: Amogh Maheshwari, Reeti Bandyopadhyay, Savir Dillikar, Spencer Liu, Raquib Alam
- Prize: $4K + Unitree G2 Pro robot dog 🐶
Notable Projects:
- Fractal — branching research spaces
- Haggler — auto-negotiates refunds
- DML — Deterministic Memory Layer for agents
10-second 720p video generation, native audio, and API that tops Artificial Analysis benchmarks
All-in-one AI video creation engine with native multimodal generation, multi-shot sequences, and built-in audio
First open-source full-duplex omni-modal LLM that can see, listen, and speak simultaneously
State-of-the-art speech-to-text, sub-200ms latency, open weights under Apache 2.0
Open-source AI music generator runs full songs in under 10 seconds on consumer GPUs with MIT license
- YouTube: thursdai.news/yt
- Newsletter: thursdai.news
- Twitter: @thursdai_pod / @altryne