Skip to content

Lead detector results with a plain label and odds - #9

Merged
ichen27 merged 1 commit into
mainfrom
feat/plainer-detector-results
Nov 15, 2026
Merged

ichen27 merged 1 commit into
mainfrom
feat/plainer-detector-results

Conversation

@ichen27

@ichen27 ichen27 commented Nov 15, 2026

Copy link
Copy Markdown
Owner

Problem

The result page led with percentiles and a zoomed scale ("scores higher than 99.6% of 223 public human texts"), which was hard to read at a glance.

Change

  • Each passage gets a label (Human-like, Slightly AI-like, AI-like, Strongly AI-like), a four-step meter and one sentence in odds: "Only about 1 in 30 human texts score this high." A human-like result also says how often AI text scores that low.
  • The whole text gets a label and a one-line reading, plus a strip with one block per passage.
  • A second opinion is reduced to its label and whether it agrees.
  • The percentile track and exact rates are kept under "Show the numbers".
  • The API and MCP results carry the same label and plain fields; existing fields are unchanged.

Labels still follow the stricter human reference and describe the score, not the author.

Validation

  • Lint clean, 51 tests pass (13 new or changed, covering odds rounding, wording and whole-text labels).
  • Rendered the page in headless Chrome with one-passage and three-passage results; the README screenshot is updated.

Risk

A label is easier to over-read than a percentile. Each label is shown with its false-alarm odds, and the limits list is unchanged.

Each passage now gets a label from Human-like to Strongly AI-like and one
sentence such as "Only about 1 in 30 human texts score this high." The
whole text gets a label and a one-line reading, with a strip showing which
passages read as AI-like. The percentile track and exact rates move under
"Show the numbers". Labels still follow the stricter human reference and
describe the score, not the author.
@ichen27
ichen27 merged commit ccca1a0 into main Nov 15, 2026
1 check passed
@ichen27
ichen27 deleted the feat/plainer-detector-results branch November 15, 2026 23:33
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