Meta, Microsoft, Apple, and Amazon report on July 29 and 30 against $724 billion in projected 2026 AI capex. Alphabet fell 7 percent despite 82 percent cloud growth after lifting capex toward $205 billion, so investors want returns, not more spend. Fortune.

The bill would hand DHS emergency shutdown authority over frontier AI systems during a national-security event. It is the first statutory emergency-stop, independent of the voluntary lab pledges due August 1. Read our take.

Opus 5 shipped in GitHub Copilot and Microsoft 365 Copilot on July 24. It leads FrontierBench v0.1 at a 43.5 percent resolution rate, so the default model in your IDE and your docs just changed under you.

The deepseek-chat and deepseek-reasoner aliases went dark on July 24 at 15:59 UTC with no extension. Any code still calling them is broken now; repoint to the V4 model IDs today. Migration notice.

If you monetize with sponsorships or a paid newsletter, pick beehiiv; if the list funnels digital products, pick Kit. Decide this week: code SRE2026 takes 20% off any annual beehiiv plan through July 31. Read it, then try beehiiv or Kit.

Claude voice now reaches Gmail and Calendar on the $20/mo Claude Pro plan, plenty for read-and-summarize. Lindy at $49.99/mo pays off the moment you need it to act unattended. Read it, or try Lindy.

We compared isolation models, cold starts, session caps, and current pricing so you can paste the call straight into your architecture doc. Cloudways is the managed pick if you want the boxes without Kubernetes. Read it, or try Cloudways.

Claude Managed Agents vs OpenAI Agents SDK vs build-your-own: for most enterprise teams a managed harness wins once you price the durable-execution substrate. Read it.

The runtime choice and the escape-cost math most teams size wrong, mapped across the path from POC to production. Read it.