Blabla Meter is a Chromium extension by maths-oliveira that scores pages against a reader’s preferences. Its local bridge sends page excerpts to Jev, then applies its own scoring formula. A separate, on-demand generative pass produces summaries and highlights.
From judgments to a score
- Jev assesses actionability, relevance, impact and depth, plus flags for clickbait and other categories.
- Application code weights the positive dimensions and multiplies in penalties to produce a score from 0 to 100.
- Page scoring uses the first 2,000 characters, so the score cannot establish the value of an entire long article.
- Section badges use separate calls; failed section calls are skipped rather than shown as scores.
Let code decide what to hide
For ambiguous page elements, the bridge asks Jev whether each is an ad and whether it disrupts reading. The extension’s thresholds determine whether to hide it. The README says obvious ad slots can also be hidden with local heuristics.
The choices of categories, weights and thresholds are product policy. A typed score is not proof of usefulness, and a false positive can hide legitimate content. Page excerpts, preferences and candidate element details go to the selected provider or TypeSafe; the local bridge does not make those calls local inference.
Verification: inspected the README and page/ad scoring code. The extension and live Jev were not run; accuracy, latency and security were not independently tested.