Stop reading AI vendors' homepages. Read their changelogs — that's where the truth ships. When I evaluate an AI tool now, I skip the homepage and open the changelog. Take Claude Code's. Multiple point

…see more

releases in a week. Version 2.1.224 alone added self-hosted runners — turn your own machines into places cloud sessions can run — plus plugin installs from plain zip archives and finer-grained credential masking for sandboxes. The three releases since are almost all fixes: terminal redraws, session resume, credential expiry. That one page tells you more than any keynote. It tells you where the product is going (infrastructure you control, not just their cloud). It tells you the team ships daily and fixes what users actually hit. And the sheer cadence tells you AI is doing a share of the work — no human-only team sustains that rhythm. This cuts both ways if you're building. Your changelog is now due diligence material. Buyers, investors and hires will read it the way I just did — as an X-ray of your priorities and your velocity. A quiet changelog under a loud homepage is a tell. The old adage was to judge people by what they do, not what they say. AI-era companies finally publish what they do, weekly, in public. Make yours worth reading. When you evaluate a vendor, what do you read first — and has AI changed it? #BuildingWithAI #ProductVelocity #Founders