AIQ Rank AIQRank
AIQ Rank · Changelog

◆ What's new

Changelog

Recent product updates, written for humans.

2026-08-30 · Product fix · plugin v0.3.28

Frozen Codex updates now finish on their own

  • What happened. If you use AIQ Rank from both Claude Code and Codex, a Claude session running an older plugin copy could quietly stall the Codex-side updater. Updates stopped halfway — the local log kept a "preserving modified" note — and the Codex scanner stayed on an old version even though the plugin itself looked current.
  • Why. The updater only replaces a local file it can vouch for: one matching the last install record, a known release, or the incoming bundle. An older plugin copy rewrote that record without the newest files' entries, so those files matched nothing and were treated as yours — left untouched, update after update.
  • What's fixed. The updater now recognizes every previously released version of those files as safe to replace, so the next update completes and the scanner catches up in one run. The install record also keeps entries for files an older copy doesn't know about, so this cannot happen again.
  • Nothing changes for your data. The same privacy bound as always — only aggregate counts leave your machine.
2026-08-26 · Product update · plugin v0.3.27

Customization now counts the setup you built

  • A setup you built months ago still counts. Customization used to measure only your edits to config files inside a rolling 30 days. That penalized exactly the wrong people: the better your setup got, the less you needed to touch it, and the lower you scored. It now measures the configuration you actually have and still use.
  • More of your environment counts. Custom subagents and custom slash commands now count alongside instruction files (CLAUDE.md, AGENTS.md) and MCP configs. Refining your config this month still helps — it is just no longer the whole picture.
  • Configuration you have stopped using counts for less. The score reflects how much of your setup you are still exercising, so it cannot be padded with files you never open. Config you have moved on from keeps partial credit rather than dropping to zero.
  • Nothing new leaves your machine. The plugin sends two numbers: how many configuration surfaces it found, and how many it saw you use. No file paths, project names, or file contents — the same privacy bound as always.
  • Nothing drops before 26 October 2026. Until then Customization takes whichever is higher, the old measurement or the new one, so updating your plugin can only help your score.
  • On 26 October the old measurement retires. From that date only the new one counts. A device still on an older plugin cannot report what you have configured, so its Customization will fall sharply — not because your setup got worse, but because we can no longer see it. Update in Claude Code or Cowork with claude plugin marketplace update aiqrank followed by claude plugin update aiqrank@aiqrank. Codex, Cursor, and OpenCode are unaffected and keep the previous measurement, since they have no way to report configuration yet. Both measurements are documented on the methodology page .
  • You will be reminded before the date, not after. If your plugin is too old to report configuration, the session-start notice now names the deadline and what it costs you, so nobody finds out by watching their score fall. Codex never shows it, since Codex is not affected.
2026-08-25 · Product fix · plugin v0.3.25

Skills you wrote on Windows now count

  • Windows authorship is credited. Skills you authored during Windows sessions were not being recognized, so they never reached your Custom Skills score. They now count the same as on macOS and Linux. Planning credit and instruction-file changes made on Windows are picked up correctly too.
2026-08-23 · Product update · plugin v0.3.24

Lighter background uploads, clearer error messages

  • Background capture uses a 7-day rolling window. The silent background uploader now sends the last 7 days of activity each session instead of a full 30-day snapshot, so it finishes faster and stays out of the way. Older days are preserved on your aiqrank.com profile; the manual full-history backfill command still covers all 30 days when you need it.
  • Rejection reasons are now readable. When the server declines an upload, the log and the manual upload command now show the parsed reason (for example, "invalid inferred role") instead of a bare HTTP code, so troubleshooting no longer requires reading the source.
  • Uploads print a link instead of opening a browser. The upload command no longer launches a browser automatically — it prints the link for you to open when ready. Pass --open to keep the old behavior.
  • Desktop and Cowork users get a nudge to add the Terminal plugin. If you use Claude Code in Desktop or Cowork but have not installed AIQ Rank in the Terminal CLI, a monthly reminder names the two commands to add it. It goes quiet once installed, and never appears in Codex.
2026-08-20 · Product fixes

More dependable from sign-in to long workdays

  • Planning scores keep their signal. An outdated measurement from one day can no longer suppress valid planning activity across your score window.
  • Activity capture keeps up with intensive days. The activity-capture guard is now relaxed so increasingly intensive Claude Code workdays stay represented.
2026-08-12 · Product update · plugin v0.3.23

Agent runtimes join the rank

  • Hermes, OpenClaw, and NanoClaw. Local activity from all three now gets its own score, profile tab, and leaderboard.
  • Automation measured honestly. Scheduled and spawned work contributes to schedule and orchestration signals without inflating ordinary session counts.
2026-08-05 · Product update · plugin v0.3.22

A quieter, smarter update path

  • No stale reminders. Once your installed plugin is current, older update notices clear instead of returning at every new session.
  • Correct Codex bundle. Codex now recognizes its Claude-compatible plugin path and uses the newest cached bundle when older managed artifacts are still present.
2026-08-04 · Product update · plugin v0.3.21

Codex speaks Codex

  • Correct update instructions. Codex now shows its namespaced AIQ Rank command when a plugin update is available, while Claude Code keeps its slash-command flow.
  • Same easy path. Run $aiqrank:aiqrank inside Codex to refresh the plugin and managed scan artifacts.
2026-08-04 · Product update · plugin v0.3.20

A clearer picture of how you work

  • Broader activity capture. Claude Code and Cowork now capture more of the command breadth and planning structure in your practice.
  • More transparent scores. Your profile now shows the methodology version behind your score, while the dimensions page explains how weighting and normalization work.
2026-08-02 · Product update · plugin v0.3.19

Codex updates itself from inside Codex

  • No copy-and-paste installer step. Run $aiqrank:aiqrank and the plugin now asks the Codex CLI to refresh the marketplace plugin and its managed scan artifacts automatically.
  • Clearer update notices. Codex uses its namespaced skill command, while Claude Code keeps its slash-command flow.
2026-07-29 · Product update · plugin v0.3.18

Scores better reflect work that lands

  • Planning that leads somewhere. Planning gets recognition when it turns into real progress, not just an outline.
  • Custom skills that make a difference. AIQ Rank better recognizes the practices you create and use to move work forward.
2026-07-26 · Product update · leaderboard at a glance

See the community from the homepage

  • Top rankings on the homepage. The homepage now gives you a quick look at the live leaderboard and the dimensions people are building strength in.
2026-07-23 · Plugin v0.3.15 & v0.3.16 · Smoother Codex updates

Keeping the plugin current got easier

  • Plugin updates apply in place. When a new version is available, the AIQ Rank plugin can bring itself up to date without sending you off to copy a command.
  • More reliable Codex activity. Recent updates make it more dependable to keep your Codex activity current.
2026-07-17 · Plugin v0.3.14 · Planning activity recognized

More of your planning practice counts

  • Native planning activity included. AIQ Rank now recognizes planning work captured directly by the tools you use.
2026-07-16 · Plugin v0.3.13 · Pi support

Pi joins the lineup

  • Track Pi with AIQ Rank. Pi joins Claude Code, Codex, Cursor, and OpenCode, so more of your AI-assisted work can live in one place.
2026-07-15 · Plugin v0.3.12 · More complete Codex activity

A clearer view of your Codex practice

  • Improved activity capture. AIQ Rank now more completely reflects the work you do in Codex.
2026-07-14 · Plugin v0.3.11 · Maintenance release

A fresh plugin release

  • Current improvements, packaged up. This maintenance release keeps the AIQ Rank plugin current.
2026-07-01 · Plugin v0.3.10 · Custom skills

Your own skills get more credit

  • Authored skills recognized. AIQ Rank now better recognizes the skills you create as part of your own practice.
2026-06-19 · Plugin v0.3.9 · Extended sessions

Long sessions stay represented

  • More reliable activity capture. AIQ Rank is more dependable at keeping your activity current across extended work sessions.
2026-06-15 · Plugin v0.3.8 · Activity detail

More useful detail in your profile

  • A clearer personal view. Your private profile can now show more detail about the AI activity behind your rank.
2026-06-10 · Product update · More of your practice counts

A fuller picture of how you work

  • More than raw activity. Recent scoring updates put more emphasis on the breadth and depth of your AI-assisted practice, so the score is a more useful reflection of how you work.
2026-05-21 · Plugin v0.3.7 · Profiles and sharing

Your rank is easier to keep and share

  • Share your profile in one click. Anyone visiting a profile can now copy its link to share an AIQ Rank with a teammate, manager, or friend.
  • Daily refreshes. Your rank stays current as new work is captured, without needing a manual refresh.
  • Project-local skills included. AIQ Rank now recognizes the custom skills you use in your own projects.
2026-05-18 · Plugin v0.3.6 · Easier setup

A friendlier path to your first rank

  • Clearer installation guidance. Getting AIQ Rank set up is simpler across the tools we support.
2026-05-12 · Plugin v0.3.5 · Team leaderboards

Compare AI-assisted practice with your team

  • Dedicated group leaderboards. Teams can compare their AI-assisted practice in a shared leaderboard.
2026-05-04 · Plugins v0.3.2–v0.3.4 · Cross-platform support

More dependable setup on every desktop

  • Windows and Mac improvements. A series of plugin releases improved setup and activity capture across both platforms.
2026-04-28 · Product update · plugin v0.3.0

See your AI practice as a whole

  • An all-tool view. Your profile and the leaderboard can now show a combined view of your supported AI tools, while still letting you drill into each one.
  • More supported tools. Cursor and OpenCode joined Claude Code, Codex, and Claude Cowork, so more of your AI-assisted work can live in one place.
  • Clearer setup. Setup guidance is now friendlier across the supported tools, helping you get from install to your first rank with less friction.
2026-04-26 · Methodology refresh

How dimensions are measured just changed

  • Renamed. context_leverage → async_delegation. Same underlying signal; clearer name.
  • Three new Claude dims. Practice Depth, Bespoke Practice, and Parallel Workstreams — all computable from data we already capture.
  • Four removed. Prompt Efficiency, Tool Diversity, Conversation Depth, and Basic Chat are no longer scored. They were either constant across users or saturated at the top of the cohort.
  • Scores rescored. Every existing score has been recomputed under the new methodology. Score history pre-this-date has been trimmed so the sparkline reflects the new normal, not a methodology-shift artifact.
  • Read more. /dimensions explains every dim in detail.
2026-04-24 · Release

AIQ Rank now ranks OpenAI Codex

  • Two leaderboards. Claude Code and OpenAI Codex each have their own leaderboard, reachable from the top nav. The "Leaderboard" word is replaced with two short links — Claude and Codex — each with a podium icon.
  • /leaderboard hub page. When you visit /leaderboard, pick which one you want.
  • One tool at a time on your profile. Your profile shows your score for one tool, with a small tab switcher when you have both.

Claude Cowork shows up on your profile

  • Cowork sessions counted. If you use Claude Cowork (the autonomous-agent mode), AIQ Rank now scans those sessions in addition to your interactive Claude Code transcripts.
  • New rows on your profile. Cowork sessions, Cowork messages, and Queue events appear in the "Raw activity" section when you have any. Hidden when you don't.

Fairer scoring per role

  • Your role sets the bar. Designers, marketers, founders and other non-engineers were graded against engineer-level usage, so they couldn't reach S-tier without behaving like engineers. Each role now has its own calibration — exceptional-for-a-marketer counts as exceptional, with proportionally less raw activity. The 1000 ceiling is reachable for every role.
  • Change your division in one click. A dropdown next to "Division" on your profile lets you switch role without leaving the page. Your score recomputes automatically.

UI polish

  • Public /changelog page. What you're reading now.
  • Cleaner leaderboard headers. Dropped the "AIQ RANK · LEADERBOARD" eyebrow and the "Ranked by AI proficiency score" lead-in. The leaderboard speaks for itself.
  • Division label on your profile. A clean role label under your handle.
  • Activity chart moved above the radar. You see your trend over time before your dimension breakdown.

Yegge developer-agent stage on your profile

  • New "Stage X.Y" chip next to your tier, mapping your activity to Steve Yegge's 8-stage developer-agent evolution model (Stage 5 "Editing Agents" through Stage 8 "Autonomous Agents" for Claude Code; 5–6 for Codex). A small "?" next to the chip links to Steve's "Welcome to Gas Town" post so you can read what each stage means.