your daily dose of claude
Meta has opened its advertising platform to outside chatbots through a new MCP server and CLI that lets Claude and ChatGPT manage campaigns, edit catalogues and pull performance reports inside the AI tool of choice. The open beta needs no developer credentials or API setup, only a standard Meta login, marking the largest ad platform yet to expose its account surface to third-party agents.
A day after the Defense Department signed AI contracts with seven rivals, Pentagon CTO Emil Michael told CNBC that Anthropic remains a designated supply chain risk and is not coming back to classified work, even as parts of the administration warm to its Mythos cyber tool. Michael framed Mythos as a separate conversation, ending speculation that the recent White House meetings signalled a thaw.
$ cat story_03.md
Counterpoint Research puts Anthropic at 31.4 percent of global LLM revenue in Q1 2026, narrowly ahead of OpenAI, despite serving roughly 134 million monthly users to OpenAI's 900 million. Anthropic earns about 16.20 dollars per active user each month, more than seven times OpenAI's figure, suggesting its premium professional pitch is paying off.
Writing in Fortune, Jeffrey Sonnenfeld and colleagues say Anthropic's Mythos preview is the wake-up call boards have been ignoring, and propose an eight-variable framework for deploying autonomous agents across banking, healthcare, retail and supply chain. They argue regulation will lag, so companies should match governance intensity to how reversible an agent's mistakes are.
The latest Claude Code release lets developers paste an auth code into the terminal when browser callbacks are blocked, fixing a long-standing pain point for SSH and container workflows. The 33-change update also stops requests from dying when a Mac wakes from sleep and patches an enforcement bypass that let lower-priority sandbox policies be silently dropped.
wotai.co →Judge Araceli Martinez-Olguin will hear final objections to the Bartz v. Anthropic settlement on May 14, with rightsholders having claimed 91.3 percent of eligible works, far above the typical 10 percent class action rate. Some authors will argue the payout is still too light a penalty for the scale of training-data piracy alleged in the case.
Anthropic is staging a second run of its Code with Claude developer conference, with sessions in San Francisco, London and Tokyo plus a livestream for everyone else. Tracks span newcomers picking up Claude Code for the first time and teams already deep into agent workflows.