Archive

Anthropic just shipped an agent that does not wait for your prompt. It reads your inbox every few hours, drafts the work, and pings you only when it needs a yes.

Five enterprise AI gateways compared on deployment, licensing, governance, failover and cost, plus a hands-on overhead and failover test.

An ICLR 2026 paper skips text between models. The most interesting number is in the ablation table.

A hands-on review of Activeloop's Hivemind: 19 skills mined from my real Claude Code history, the pattern it refused to write, and the recall bug I traced and fixed.

The hard part of agent evals is the sandbox.