Clear Thinker
DailyWrite 5 prompts that make clear what you want (8 or more) today.
An RPG for your Claude Code sessions
Path of Dev grades each prompt you send against what it needs (a clear ask, the right pointers, a picture of done) and turns the grade into XP, loot and a pixel pet that grows with you. Shipping real work on GitHub pays the big bonuses.
› parseDue fails on DST days in src/dates.ts. Fix the offset math so the date tests pass again, and don't touch the public API. ✦ Graded 8.9/10 +13 XP intent 9 · grounding 9 · done 9 · scope 9 “clear target, testable finish” ✦ Loot: Knit Scarf [uncommon]
╡ ⚔ Lv 7 · Apprentice ╞════════════◇═══ █████████░░░░░░░░░░░ 2,793 / 3,155 ● 412
A small model grades every prompt of six words or more against what that prompt actually needs: six measures of what it says, each from 0 to 10, and a little more for clear writing and a kind tone. A quick follow-up question isn't marked down for skipping a why it doesn't need.
parseDue fails on DST days in src/dates.ts. Fix the offset math so the date tests pass again, and don't touch the public API.
Grader's note: “clear target, testable finish”
Your XP is your grade times one and a half, so a perfect 10 earns 15. Prompts are where the daily XP comes from.
Spamming prompts doesn't pay. The rate drops as the day goes on.
Grading runs inside your own Claude Code. The game server receives only the scores, and whether the prompt was spelled clean. Your prompt text, the grader's note and any word it flagged never reach it.
Prompts are the everyday XP. Real outcomes pay much more, and daily limits keep busywork from adding up: a hundred tiny commits are worth 60 XP.
| What you did | Daily limit | XP |
|---|---|---|
| A graded promptyour grade × 1.5 | tapers after 20 | up to 15 |
| First contribution to a repo GitHub | none | 300 |
| Pull request merged120 on your own repo | none | 250 |
| Bug fixed GitHub | 5 a day | 150 |
| Review feedback acted on GitHub | 10 a day | 80 |
| Issue closed GitHub | 5 a day | 60 |
| New repo explored | 3 a day | 25 |
| Red tests turned green | 5 a day | 20 |
| Streak day | 1 a day | 10 |
| Commit | 20 a day | 3 |
| Pull request openedit pays out when it merges | none | 0 |
Rows marked GitHub are checked against GitHub before they count. Only confirmed work counts toward the leaderboards.
Each turn can drop an item, 12% of the time, and each graded prompt rolls too: 17.5% of the time at a perfect 10, less as its grade falls. A grade of 9 or more doubles the odds of rare and better. Every level-up drops one too.
“It hums when the tests go green.”
“Said to glow brighter with every merge.”
Bad luck runs out: a rare is guaranteed within 25 drops, and an epic within 80. Wear your loot on your pet or use it to restyle the band above your prompt.
Three dailies every day. Each week, five weeklies are on offer and you run three of them.
Write 5 prompts that make clear what you want (8 or more) today.
Write 3 prompts that say what done looks like (7 or more) today.
Get 25 prompts graded very well this week.
Get a pull request merged this week.
Others ask you to turn a red test suite green, keep context under half for three sessions, or run a code review three times in a week.
Your title changes as you level, and your pet evolves three times along the way. Each level takes a little more XP than the last.