Skip to content

docs(readme): give your agent a program, not a mouse - #1273

Merged
abrichr merged 5 commits into
mainfrom
readme-compile-and-check-2026-10-18
Oct 19, 2026
Merged

abrichr merged 5 commits into
mainfrom
readme-compile-and-check-2026-10-18

Conversation

@abrichr

@abrichr abrichr commented Oct 18, 2026 •

Copy link
Copy Markdown
Member

Reframes the README as the harness for computer-use agents. Stacked on #1260.

  • Headline: "Give your agent a program, not a mouse." The subhead is "The open-source harness that compiles computer-use agent runs into verified programs."
  • One harness, three modes: explore (an agent or a person does it once), run (0 model calls, record checked before and read back after), repair (it stops before saving; an agent shows the fix and a person approves it).
  • An agent alone vs an OpenAdapt program: the 62-run OpenEMR table. Same correctness, $0.08 vs $0, screenshots vs none, 36.5 s vs 29.7 s, with the honest note on why the speedup is small and the API-step PR (feat(flow): API steps compiled from a recording's network capture #1269).
  • Then: the fault-test table, the quickstart, and "Build with it" with the agent path first (compile your agent's run, call programs over MCP or Agent Skills, record by hand, judge or train any agent).
  • Every backend: the surfaces table is visible. The agent-compile path is labelled as web today, with native, RDP and Citrix in progress.

Not used: invented figures from outside reviews, and "crystallizing". README tests pass (21).

🤖 Generated with Claude Code

abrichr and others added 3 commits October 18, 2026 13:53
Open with the transformation in plain words: a person's recording or an
agent's run compiles into a program that checks every save. Put the
fault-test result first, show three ways to start (a recording, an agent's
run, underneath an agent), add an agent-with-a-replay-cache column to the
comparison, and say when to use something else.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Open with "Give your agent a program, not a mouse": an agent or a person
does a task once, the run compiles to a program, and the agent calls it as
one tool. Show the three modes (explore, run, repair), the agent-alone versus
program table from the OpenEMR runs, the build-with-it commands with the
agent path first, and every backend in the open.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@abrichr abrichr changed the title docs(readme): lead with compile, and check every save docs(readme): give your agent a program, not a mouse Oct 18, 2026
abrichr and others added 2 commits October 18, 2026 15:06
Say what was injected (8 kinds of bad save, 9 runs each) so the counts
read as a coverage test, not an error rate. Call OpenAdapt the verified
execution layer for computer-use agents, and move the timing table below
Build with it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@abrichr
abrichr merged commit 6fd19b6 into main Oct 19, 2026
36 checks passed
@abrichr
abrichr deleted the readme-compile-and-check-2026-10-18 branch October 19, 2026 00:53
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant