Jason Says
The wildest moment today: OpenAI announced an AI solved a $1M Millennium Prize problem, and a mathematician immediately fired back claiming credit was stolen — when AI capabilities start triggering attribution disputes at this level, that itself is the real milestone.
OpenAI published an AI-generated solution to the Navier-Stokes Millennium Prize Problem with a formal Lean proof, but an NYU mathematician publicly accused OpenAI of foul play, claiming the problem was central to a researcher's career-making work. The $1M bounty has ignited a fierce debate over credit and attribution between AI labs and human mathematicians.
OpenAI launched ChatGPT Images 2.5, which better translates ideas, sketches, and reference photos into polished, personalized images. The upgrade significantly narrows the gap between creative intent and output quality, making it a meaningful tool upgrade for designers and content creators.
An MIT researcher demonstrated GPT-5.6 Sol with Codex autonomously running quantum computing experiments, analyzing results, and calibrating qubits. This is one of the most concrete real-world examples of AI agents conducting genuine scientific work, signaling that AI-driven research is moving from concept to practice.
Meta launched Muse, a personal AI agent requesting access to users' email, calendars, payments, and health services. It's Meta's biggest consumer AI bet yet, but also its riskiest — directly testing whether users still trust the company with intimate personal data after years of privacy controversies.
Cognition's valuation hit $48B with a multiple exceeding Cursor's pre-SpaceX-acquisition level. Investors are betting that AI coding is a multi-winner market — not a monopoly — which is encouraging news for developers and startups building in adjacent coding-AI niches.
A Claude user discovered their account was consuming tokens without any activity, prompting Anthropic to warn users about hackers stealing Claude tokens. Developers relying on Claude API should treat token security — including key rotation and usage monitoring — as an immediate priority.
context-mode optimizes context windows for AI coding agents by sandboxing MCP tool outputs (98% reduction — e.g., Playwright snapshots shrink from 56KB), persisting session memory, and enforcing routing across 17 platforms. A practical must-have for Claude Code or Cursor users constantly hitting context limits.
The U.S. Department of Defense issued an official advisory warning that Chinese AI companies are conducting systematic malicious distillation campaigns against U.S. AI models. This is the first official designation of model distillation as a national security threat, likely accelerating API access controls and export policy tightening.
Investors: Samsung Electronics
欧洲最大开源大模型公司，估值近翻倍至 240 亿美元，三星领投彰显亚欧 AI 产业联盟趋势加速。
Subscribe for daily AI updates + free playbook
📘 Subscribe Free