On August 11, Anthropic expanded its Compliance API for auditing so that session records from Claude Code and Cowork running on employees' machines can be read out. Agent conversations that once stayed local now fall within the scope of corporate audits — here's how it works, and the tension between
Claude Code v2.1.228 is out. It closes gaps in "convenience coming from outside"—barring synced cloud skills from overreaching, stopping remote-resume from peeking at conversations, and preventing cleanup from deleting memory—while loosening the Write tool by one notch to allow overwriting unread fi
Claude Code v2.1.227 is mostly bug fixes. But it closes two holes that are easy to overlook in unattended operation: one that wrongly nudged Max users toward pay-as-you-go billing via an expired token, and one that wiped out every Bash command in CI on GitHub runners. Here's what to double-check.
On August 14, Claude Code switches the default for Pro, Max, and Team from manual approval to auto mode. We break down the design that stops only dangerous operations, the 89%-vs-13.6% detection rates, and the lingering concerns over the remaining 11% and supply chain attacks.
Cloudflare has released Kitesurf, a browser built exclusively for AI agents. It drops human-facing features and reads the web using up to one-seventh the memory of Chromium. We break down its practical usability with Puppeteer/Playwright/MCP support, along with the concerns that remain around concen
OpenAI halted some work on its next model, "Astra." For the first time, it could not rule out that a model of its own might reach "Critical" — the highest danger level in its safety framework. We unpack the backdrop of agentic coding and cyber capabilities rising in tandem, and what it means for dev
On August 6, Anthropic added a malicious-content review (Enterprise, beta) targeting third-party Claude Code skills and plugins. Behind it are researchers' warnings that extension marketplaces can become a supply-chain attack surface. Here's how it works and what adopters can do to protect themselve
Claude Code v2.1.225, released on August 8, focused intensively on "silent failure" bugs that break unattended, headless runs — 401s from expired tokens, broken authentication on macOS, and more. We look at this update, which also adds a gateway spend-limit display, from an operational perspective.
Codex CLI 0.147.0 became the first major agent to support "Agent Plugins," a vendor-neutral shared standard. From a developer's perspective, we lay out what the standard does to kill the hassle of rebuilding extensions per format, how it's governed, and what Claude's absence means.
Claude Code v2.1.224 adds a "self-hosted runner," letting web, mobile, and desktop cloud sessions run on your own machines and containers. The release also brings inter-session messaging and a partial rollback of the subagent caps Anthropic kept tightening through July.
On August 6, Cursor detailed how its Router "chooses which model to use." We break down the mechanism—Compass gauges how easy a turn is, then work is split by role (Git=Grok, planning=Sol, execution=Opus, debugging=Fable)—along with the published cost-savings figures and the caveats around reproduci
Claude Code v2.1.223 (August 6) fixes, in one release, a hole that let workflows escape the sandbox via dynamic import() and the memory leaks that had been accumulating in long-running, unattended sessions. Here's a rundown of an update that matters most to anyone running automation.
