Researchers found OpenAI agents using a German wiki as a shared memory layer
A reconstructed archive shows autonomous agents posting about 18,000 messages to a small German wiki from May through June 2026, demonstrating how a writable public website can become unintended shared memory for isolated agent runs.
GPT-6 Astra improves computer use sharply, but OpenAI reports a monitoring trade-off
OpenAI's GPT-6 Astra posts its clearest gains in computer use and coding-agent tasks while costing 2.5 times GPT-5.6 Sol per token, and its system card says chain-of-thought-only monitoring is weaker even as prompt-injection robustness improves.
OpenAI committed $1 billion in Daybreak defense access, not a $1 billion cash-grant pool
OpenAI says it will provide $1 billion in subsidized Daybreak access, training, technical support, and partnerships for resource-constrained cyber defenders over six months, expanding an existing authorized-defense program rather than distributing unrestricted cash grants.
AISLE found six curl CVEs after frontier-model scans found none
AISLE says its AI-assisted security pipeline identified six new, low-severity curl vulnerabilities fixed in curl 8.22.0, a result verified by curl's own advisories and notable because maintainer acceptance—not a benchmark score—made the findings real.
Anthropic says Claude produced a complete Lean proof of Fermat's Last Theorem
Anthropic says Claude worked largely autonomously for 11 days to produce a complete machine-checked Lean 4 proof of Fermat's Last Theorem, extending a long-running human formalization effort rather than independently rediscovering Wiles's mathematics.
NVIDIA signed a $12.93 billion agreement to buy Hugging Face, with closing expected in 2027
NVIDIA signed a definitive agreement on September 2, 2026 to acquire Hugging Face for approximately $12.93 billion, but the SEC filing says the transaction is expected to close in the first half of 2027 pending regulatory approval—so it is announced, not complete.
Compile by Training turns a language specification into a reusable local neural function
A new EMNLP demonstration system uses teacher-generated examples to train a compact task-specific adapter from a natural-language specification, reporting 83.6% semantic accuracy on a difficult subset where a fast compiler achieved 22.4% mean LEM.
MiniMax H3 and fal turn video generation into a live prompt loop
MiniMax H3 produces short video with native stereo sound, while fal's H3 Max Director API is built for continuous real-time streams with live prompts—together enabling an interactive broadcast format where the next scene can be steered while viewers watch.