bradtraversy.dev — 2026-08-28-ai-blueprint-1-1-browser-tests.md
home.md projects/ tools/ devlog/ × articles/ now.md about.md
2026-08-28 · #ai-blueprint · #devlog #agents #agent-workflow #tool-launch

# ai blueprint 1.1 adds browser tests

links: site | repo | project page

AI Blueprint 1.1 is about making the workflow useful for apps you can actually see, click, and break. The new /browser-tests command adds an optional browser automation gate to a project, records the command in the local config, and lets Feature, Check, and Continuous Mode use that proof when UI behavior matters.

The release also adds project-level automation policy. A project can say how much help it wants from regular commands versus Continuous Mode, and the default still stays conservative: audit, check, and try work remain manual unless the project opts into more.

What shipped:

  • create-ai-blueprint@1.1.0 on npm
  • browser-test setup for Playwright-friendly JavaScript and TypeScript apps
  • Continuous Mode that works locally, one feature at a time, without pushing or publishing anything
  • a redesigned read-only dashboard over repo state and blueprint/.state/run.json
  • sandbox coverage for onboarding, adapters, CLI behavior, and installer paths

I also tightened the docs around where browser automation fits. Unit tests prove logic. Build checks prove the project still compiles. Browser tests prove the parts that humans actually notice: visual states, auth paths, extension UI, interactive flows, and regressions that only show up after a click.

That is the direction I want AI Blueprint to keep moving: more proof, more project memory, and more automation only where the project has explicitly asked for it.

// EOF 2026-08-28-ai-blueprint-1-1-browser-tests.md
main
2026-08-28-ai-blueprint-1-1-browser-tests.md
UTF-8
LF
Markdown
Ln 1, Col 1