Benchmark Radar Day 38: A Blog Built From Evidence

Day thirty-eight of Benchmark Radar. The daily brief, the thing the project already wrote every day and never gave a page, now has a blog and a feed. Scoreboard: 137 stars, 26 forks.

Author: Koutian Wu; GitHub: ktwu01

Benchmark Radar is live. Track new AI benchmarks, datasets, and leaderboards every day: open the dashboard or star it on GitHub.

A blog is a set of dated pages, one per day. An RSS feed is a machine-readable list that readers and apps subscribe to. A snapshot is the stored record of one day’s collection. A briefing is the short daily summary of what moved.

PR #476 gives the daily brief a home. The briefings, questions, caveats, and coverage numbers that were written into every snapshot and never displayed now publish at /blog/, with a landing page, a full archive, and an RSS feed. The leaderboard, trends, and explore views return real HTML in the same visual language, and translated posts use one canonical URL with a same-page toggle.

PR #494 makes a briefing source directly clickable. When the day’s evidence resolves to exactly one validated citation, the single source chip becomes a link to it; multi-source chips stay as counts rather than ambiguous links.

PR #507 ends every query CLI round with a citation reminder. A single source file mirrors the citation file and the version number, so a release bumps the citation without a second edit, and in machine mode the reminder rides on the error stream so the data stays clean.

PR #505 fixes Today rendering empty after a sheet is closed or Back is pressed, and PR #509 pushes the contributors image straight to main using an app token and a path allowlist instead of an unreviewed write. PR #517 reduces the Agent Skill setup to one command and lets the agent detect and repair a broken CLI itself.

Why this matters.

Evidence that is written but never published might as well not exist. A blog gives the daily record a URL, a feed gives it subscribers, and both are built from the snapshot alone, with no model call at build time.

A citable CLI is part of the scientific loop. Ending every query round with a citation ask is a small nudge that compounds every time someone uses the tool.

Issues addressed

• #448: evidence blog and static pages while preserving the dashboard UI
• #467: singular briefing sources directly clickable
• #483: every CLI round ends with a citation reminder
• #503: Today renders empty after closing a sheet or pressing Back, fixed
• #516: Agent Skill setup reduced to one command
• scoreboard: 137 of 1,000 stars, 26 forks

Day thirty-nine: the daily brief gets a landing page, an archive, and a feed of its own.

Want to follow Benchmark Radar? Star the repo on GitHub for daily updates, or open the live dashboard to explore the scans.