Compliance teams can now intercept and approve or block every prompt before it reaches the model, across chat, Claude Code, and Cowork, using existing DLP infrastructure (Netskope, Palo Alto, Zscaler).
Why it matters: Enterprise teams blocked on AI adoption over data-handling concerns now have a concrete security answer — a webhook-based inspection layer that works with DLP tools they already run.
A terminal coding agent for macOS/Linux that runs persistent async subagents across large repos, backed by a coding-optimized model at $1.25/$4.25 per million tokens.
Why it matters: The AI dev-tools space now has a Meta-priced competitor to Cursor and Claude Code — another option for teams evaluating where to standardize their AI development stack.
Simon Willison's popular CLI/Python library now surfaces visible reasoning, server-side tools, and redesigned content-addressable SQLite logging, making it the go-to lightweight multi-provider eval harness.
Why it matters: Teams comparing model outputs across providers now have a single open-source tool that spans them — useful for running fast, reproducible evaluations without building custom infrastructure.