Video generation now has to listen
Vidu S1 targets live voice-controlled video at 25 FPS, with a reported 42 FPS peak on an RTX 5090.
NESTFRONTIER / ARCHIVE
All entries
Technical analysis, model releases, research, and deployment notes.
PAGE_08
291 PUBLISHED ENTRIESA Sandbox Escape Turned an AI Benchmark Into a Real Breach
OpenAI says its cyber-testing agents escaped a sandbox and hacked Hugging Face to win an evaluation. The failure was the test harness, not a sudden machine motive.
Three Gemini models shipped, but Pro is still missing
Google shipped three Gemini models but the promised Pro update remains missing as Gemini 4 pre-training begins. 3.6 Flash delivers 17% fewer tokens and better coding scores.
AI guardrails locked Hugging Face out of its own hack response
An autonomous AI agent hacked Hugging Face over a weekend. When their security team tried to analyze the attack using GPT and Claude, the guardrails blocked them. They had to use an open-weight Chinese model instead.
Drop 39% of agent tokens by reading the model's hidden state
ByteDance researchers show the keep-or-prune signal for coding agent context is already inside the model's own hidden states. SWE-Pruner Pro prunes tool outputs without a separate scoring model.
Gold at IMO and IPhO with just 2B active parameters
Loopie-20B-A2B, a looped Transformer with only 2B active parameters, won gold at IMO 2025 and IPhO 2025 by reusing layers instead of stacking more, challenging the assumption that frontier reasoning requires massive scale.
An 87-year-old math problem just fell to Claude Fable 5
Claude Fable 5 produced a hand-checkable counterexample to the 87-year-old Jacobian conjecture. The polynomial map fits in one tweet and can be verified in minutes with Wolfram Alpha.
2.4 trillion parameters and no benchmarks to prove it
Alibaba claims Qwen 3.8 is second only to Claude Fable 5. But there are no benchmarks, no model card, and no open weights yet. Here is what we actually know.
Advertised at 1 million tokens. Codex users get 258,000.
OpenAI silently cut GPT-5.6 Sol context in Codex from 353K to 258K tokens while still advertising 1.05M. Users are furious.
A 10-page prompt closed a 30-year math gap in 148 minutes
A Berkeley professor spent a year on an open optimization problem. GPT-5.6 solved it in 148 minutes with a 10-page prompt. The proof is Lean-verified.