Anthropic may have degraded its developer experience to hide a capacity crisis. Over the past several weeks, the company shipped a noticeably weaker model and revoked Claude Code access from paying accounts. The timing matters: Anthropic recently secured a major compute deal with SpaceX. The question the original piece asks directly is whether the hostility was a symptom of constrained infrastructure, not a product decision.
Elsewhere in the piece: Amazon has lifted its internal ban on Claude Code and Codex after pushing engineers toward its own Kiro coding agent. Meta is forcing 20-40% of engineers on certain teams into data labeling roles ahead of expected cuts, which may temporarily protect headcount. Both Meta and Amazon CEOs are publicly endorsing 5-10 person AI-forward teams over teams of 50, with no clear answer yet on what happens to the people who don't make that cut.
The GitHub Copilot pricing update buried in the opening is worth your attention. Annual renewals will see roughly a 3x cost increase. Pro, Pro+, Business, and Enterprise plans are moving from request-based to API token-based pricing on June 1, making costs unpredictable from day one. The full piece also covers Meta's employee computer monitoring, OpenAI migrating off Datadog, Apple quietly confirming internal Claude Code use, and a VS Code bug that tagged commits as Copilot-authored when Copilot did nothing.
[READ ORIGINAL →]