AI infrastructure is attracting multibillion-dollar cloud commitments and financing.
$ conclusions
> the model raceHow the top models compare. A lab counts as ahead only when independent, comparable tests show it.
Google has begun limited access to Gemini 4 Argon for cyber defenders; Anthropic’s Opus 5.5 remains ahead of Sonnet 5.5 on one independent index, but no overall leader is established.Since Sep 30
> what's changingThe strongest change the evidence supports so far.
Local-model support is appearing in agent products.early signalSince Sep 26
> what's settledA conclusion with enough evidence to treat as lasting.
No durable conclusion is established yet.
$ since-last-update
OpenAI now offers optional text watermarking in its API, and the Pentagon says it has stopped using Anthropic tools. what changed on Oct 5
Researcher Syed Anas Mohiuddin documented request-forgery vulnerabilities that he says five organizations confirmed and fixed, including Google and JPMorgan Chase. An agent-supplied destination could direct an MCP server's outbound request to an internal endpoint; Ars Technica…