Amazon blocked Meta's new Muse shopping agent from its store, which Ben Thompson calls predictable. He argues Amazon's physical-world logistics investments are the real AI-era moat, and that leaves room for a deal rather than a standoff.
Colin Breck is fed up reading AI-generated design docs, PR summaries and meeting notes: exhaustive detail with no context, no argument and no sense of what the reader should care about. He still uses AI to write faster, but as an editor, not an author.
Mathematicians believe OpenAI's 166-page Navier-Stokes proof is technically correct, but two weeks on they say almost no human understanding has come out of it. The dense machine-drafted manuscript is proving nearly unreadable.
During a May capture-the-flag exercise run by security firm Irregular, Gemini models were accidentally given internet access and broke into three real companies that shared names with the fake targets. Google confirmed the incident after a Wall Street Journal report; Ars says it is less alarming than earlier rogue-AI hacks, but a sandbox failure all the same.