AI coding efficiency · public leaderboard

Tiny tokens.
Big ships.

Track your AI coding burn, compete in verified cost-per-fix challenges, and learn the workflows behind top builders.

Claude CodeCodexCursor+ more

Top VES201.7

Week fixes3,391

AI spend$4,820

Burn Index · Live

#1@shellcoder201.7$4.20

#2@tinyshipper197.8$3.60

#3@noor195.5$2.20

🏆 @shellcoder · VES 201.7 · Provider-synced⚡ @tinyshipper · $3.60 / 712 fixes · Device-synced🔥 Challenge live: Fix a Lighthouse regression — $50 reward📦 New drop: Claude Code Review Loop by @tinyshipper✅ @noor · VES 195.5 · +2% this week💡 Workflow drop: Low-Cost Debugging — Free🏆 @4ndres · 378 fixes · Provider-synced⚡ Week total: 2.4B tokens · $4,820 AI spend · 3,391 fixes🏆 @shellcoder · VES 201.7 · Provider-synced⚡ @tinyshipper · $3.60 / 712 fixes · Device-synced🔥 Challenge live: Fix a Lighthouse regression — $50 reward📦 New drop: Claude Code Review Loop by @tinyshipper✅ @noor · VES 195.5 · +2% this week💡 Workflow drop: Low-Cost Debugging — Free🏆 @4ndres · 378 fixes · Provider-synced⚡ Week total: 2.4B tokens · $4,820 AI spend · 3,391 fixes

Burn Index

Who ships the most for the least?

Ranked by VES — Verified Efficiency Score (verified fixes ÷ AI cost USD). Lower spend, more fixes = higher rank.

#BuilderVerificationTokensCostFixesVESTrendSpark

Verified4

Provider or device-synced — measured at the source.

@shellcoderDevice-synced2.1M$4.20847201.7+12%

@tinyshipperProvider-synced1.8M$3.60712197.8+8%

@noorDevice-synced1.1M$2.20430195.5+2%

@4ndresProvider-synced980K$1.96378192.9+1%

Self-reported1

Submitted by the builder, not yet confirmed.

@coconutfixSelf-reported3.2M$9.601,024106.7-3%

VES = verified fixes ÷ AI cost (USD). Higher is better. Verification badges: Provider-synced > Device-synced > Estimated > Self-reported.

Cost-per-Fix Challenge

Compete on verified efficiency.

Every challenge is judged by VES — independently verified fixes divided by your total AI spend. No vanity metrics.

$50 reward48h left

Fix a real Lighthouse regression

Ship a verifiable fix that improves Lighthouse score by ≥10 points. Claude Code only.

14 participants

$30 reward72h left

Zero-token refactor

Refactor a component with the lowest possible token spend. VES judged.

9 participants

$20 reward24h left

Bug hunt: cost < $0.10

Find and fix a verified bug for under 10 cents of AI spend.

22 participants

3Active challenges

$100Total reward pool

45Participants

score.py

# Verified Efficiency Score
def score(fixes: int, cost: float) -> float:
    return fixes / cost

# Example from this week
score(fixes=847, cost=4.20) # → 201.7

Builder Cards

The builders setting the pace.

Each card shows verified efficiency over the last 7 days. Badges reflect how usage data was collected.

@shellcoderDevice-synced

201.7VES

847fixes

$4.20spent

+12%#1 this week

@tinyshipperProvider-synced

197.8VES

712fixes

$3.60spent

+8%#2 this week

@coconutfixSelf-reported

106.7VES

1,024fixes

$9.60spent

-3%#3 this week

@noorDevice-synced

195.5VES

430fixes

$2.20spent

+2%#4 this week

@4ndresProvider-synced

192.9VES

378fixes

$1.96spent

+1%#5 this week

Live activity

@shellcoderfixed 12 bugs$0.482m ago

@tinyshippercompleted challenge$1.2014m ago

@noorfixed 8 bugs$0.3231m ago

@4ndresnew workflow drop—1h ago

@coconutfixfixed 24 bugs$2.162h ago

Workflow Drops

Proven workflows from top builders.

Read-only guides, config packs, and executable bootstraps — all verified by VES score. No black-box prompts.

readonlyFree

Low-Cost Debugging

A step-by-step workflow for debugging with minimal token spend. Read-only guide.

@shellcoder

config$24

Claude Code Review Loop

Pre-configured review loop that cuts review token cost by ~40%.

@tinyshipper

executablesoon

Codex Repo Bootstrap Pack

Executable scaffold pack for starting new repos with Codex-optimized structure.

@4ndres

Trust & Safety

Built for builders who own their data.

We track efficiency metrics — never your code, prompts, or secrets. The collection spec is public and auditable.

No raw prompt storage

We never store, process, or transmit your source code, prompts, or responses. Zero.

Device-local collection

Usage data is aggregated on your machine. Only token counts and cost estimates leave your device.

Transparent verification

Every badge shows exactly how it was verified: Provider-synced, Device-synced, Estimated, or Self-reported.

Open collection spec

The collector spec is public. Anyone can audit what fields are collected and what stays local.

CoconutLabs never stores raw prompts, source code, or file paths. Only aggregated efficiency signals leave your device. Read the full collection spec →

Start shipping smarter today.

Join 1,247 builders already tracking their AI coding efficiency. Free to join. No credit card.

Invite-only beta · Claude Code, Codex, Cursor supported

Tiny tokens.Big ships.