Skip to content

0.7.1 — the version it reports, and the page teachers land on - #86

Merged
peopleworks merged 1 commit into
mainfrom
release-0.7.1
Sep 7, 2026
Merged

peopleworks merged 1 commit into
mainfrom
release-0.7.1

Conversation

@peopleworks

Copy link
Copy Markdown
Owner

Two corrections that never reached anyone from 0.7.0, both about what this project says about itself rather than what it measures.

#84 signsofai --version reported 0.1.0 for seven releases. It now reads the assembly — and this release is the first time that is visible: the tool built from these files answers 0.7.1, which is what the csproj packs.
#85 why.html, the page README.md:15 sends teachers to first, was still showing ninety texts and a green zero six days after the published rate stopped being zero.

The guard earned its keep one day later

Bumping the version made the calibration declare 0.7.1, and WhyPageTests failed by name:

SignsOfAI.Core.Tests.WhyPageTests.The_page_names_the_engine_that_produced_its_figures [FAIL]

That is the whole argument for a guard over a comment, demonstrated on its first real occasion. In August the same page carried a comment asking the next person to remember, and the next person did not.

Calibration

Re-run over all 296 texts. Only the engine field moved — threshold 30, 2 of 296 flagged, rateLow, rateHigh, every language and group row, and the corpus fingerprint 78bda061bde3dc99 are unchanged. There are no score changes in this release: Core's only edits are comments.

⚠️ Order of operations

DesktopRelease.Version is 0.7.1 and the workflow refuses a desktop-v* tag that disagrees with it.

  1. gh release create desktop-v0.7.1 --target release-0.7.1 — from this branch, before merging
  2. merge this PR
  3. gh release create v0.7.1 → NuGet publishes
  4. wait for NuGet to index
  5. only then the MCP registry, which is manual and separate

432 tests.

🤖 Generated with Claude Code

https://claude.ai/code/session_015PEbbiYSNPw7jE3LrPNhyF

Two corrections that never reached anyone from 0.7.0, both about what this
project says about itself rather than what it measures.

`signsofai --version` reported 0.1.0 for seven releases (#84). It now reads the
assembly, and this release is the first time that is visible: the tool built
from these files answers 0.7.1, which is what the csproj packs.

`why.html` — the page README line 15 sends teachers to first — was still
showing ninety texts and a green zero six days after the published rate stopped
being zero (#85). It now carries 296, two flagged, a bound of 2.4%, and the
paragraph that inverted: 206 of the corpus are classroom essays by adults
learning English, which is the group the boundary moved to protect.

The guard added with that page earned its keep immediately. Bumping the version
made the calibration declare 0.7.1, and WhyPageTests failed by name —
`The_page_names_the_engine_that_produced_its_figures` — instead of letting the
page go stale in silence the way it did in August. That is the whole argument
for a guard over a comment, demonstrated one day later.

Calibration re-run: only the `engine` field moved. Threshold 30, 2 of 296, both
bounds, every language and group row, the corpus fingerprint — unchanged. No
score changes in this release; `Core`'s only edits are comments.

`DesktopRelease.Version` is 0.7.1, so `desktop-v0.7.1` has to be cut from this
branch before it merges, or the download button points at a 404.

432 tests.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015PEbbiYSNPw7jE3LrPNhyF
@peopleworks
peopleworks merged commit 53787c6 into main Sep 7, 2026
6 checks passed
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