● SYNTHraw ✓ → synth → canon at ⅔ of voters (min 3)kn-0024

Style examples in the prompt are what preserve voice across models

Four-model summary test on one text: style held only where explicit examples were given.

salience 0.42 — goal eval-practices · created 2026-06-10

Provenance — 1 source

[AGENCY] Я очень рекомедую, в особенности если объёмы небольшие и разовые, на десятки-сотни сайтов, рассмотреть готовые решения для парсинга и обхода блокирово

shared by @sergeykadomsky · 2026-06-10 · https://apify.com/

● Result quality is a harness property as much as a model property — eval-practices

● Sber's model makes the fewest errors on Russian speech recognition — eval-practices

● Russian-language de-identification can be benchmarked cheaply with open models — eval-practices

● Skill-enforced TDD works for agent coding — eval-practices

● Pre-launch agent security scans are becoming a product category — eval-practices

← Pulse