The social layer for AI builders

Don’t count your tokens.
Make your tokens count.

Run in your terminal: npx ai-overflow

AI OVERFLOW / LIVE WORKSPACECodeburn connected
Explore the product
Analyze the workTrack the work, not a vanity score.Tokens, cost, time, and tool stay together.
Two honest scores
70%Efficiency
30%Burn
Score model

70/30 efficiency/burn

Private report detail

Comparable all-time signals

Rank with context

Compare completed reports.

Switch between verified token burn, cost, streak, and efficiency—not a single pay-to-win score.

TokensCostStreakEfficiency
Share + verify

Turn the session into proof.

  1. 1Share a burn card or useful answer.
  2. 2Another builder confirms the outcome.
  3. 3Credibility is awarded once.
Connect from your terminal
$ npx ai-overflow ✓ Codeburn report collected ✓ Report uploaded Open AI Overflow to see your rank.
Built for your stack41 integrations
  • Codex
  • Cursor
  • Claude Code
  • Copilot
  • Gemini CLI
  • Zed
+35 more
CodexCursorClaude CodeCopilotGemini CLIZed
41supported tools
70/30efficiency / burn
RAWreport-level detail
PEERverified outcomes
WHAT WE TRACK

Six signals. Two honest scores.

First-principles measures of how much AI you use—and whether that usage turns into useful work.

S1

Token burn

Total input and output tokens, normalized across your tools and models.

Burn
S2
$

Spend

Actual model cost, tracked from completed reports without hiding expensive sessions.

Burn
S3

Consistency

Session time, active days, and the streak you kept alive while shipping.

Burn
S4

Useful answers

Technical answers that other builders accepted because they solved the problem.

Credibility
S5

Shipped outcomes

Projects, fixes, and deploys connected to the session that helped create them.

Credibility
S6

Verification

Results confirmed by somebody other than the author—no self-awarded reputation.

Credibility
02 / THE WORKFLOW

Connect it. Make it count.

Three deliberate moves from private work to a signal that holds up in public.

01

Connect the work

Bring in supported tools or log a session directly. Your stack stays yours.

02

Read the report

Keep tokens, cost, time, model, and tool together instead of flattening the story into one score.

03

Share proof

Post a useful answer or outcome. Credibility is awarded only when another builder verifies it.

03 / ONE LEDGER

Every agent.
One clear record.

Use Codeburn-connected tools without changing how you work. AI Overflow keeps the detail readable after the session ends.

CONNECTED STACKBring the tools you already use.
  • Codex
  • Cursor
  • Claude Code
  • Copilot
  • Gemini CLI
  • Zed
  • + 35 more
WHERE THE DATA COMES FROM

Real numbers.
Shared by you.

No mystery score and no scraped activity. Usage comes from connected tools or sessions you log; credibility comes from public actions and verified outcomes.

01

User-connected or logged usage

Import supported agent usage or enter a session directly with the tool and model attached.

02

Raw metrics preserved

Tokens, cost, and minutes stay separate so every leaderboard can explain its ranking.

03

Peer-verified outcomes

Credibility points require an accepted answer or a result confirmed by someone else.

04 / YOUR USAGE, EXPLAINED

From your terminal
to the rankings.

Start with your local coding history. Preview your usage, connect your account, and see how your work compares.

Burn measures usage.Efficiency rewards useful work.Credibility needs proof.
See the full leaderboard
ai-overflow — example session
$ npx ai-overflow✓ Codeburn report collected✓ Private paths removed✓ Efficiency scored (70/30)✓ Report uploaded→ Open AI Overflow to see your rank
RankBuilderBurnSpendStreak
Loading verified ranks…
FAQ

Good questions.
Clear answers.

Does burning more mean I’m a better developer?

No. Burn is a usage metric, not a skill score. Credibility is earned separately through useful answers and verified outcomes.

Where does usage data come from?

From supported tool connections and sessions you choose to log. Every burn keeps its tool, model, tokens, cost, and duration attached.

Can I rank with a smaller budget?

Yes. Token, cost, streak, and efficiency boards expose different kinds of verified usage instead of one pay-to-win number.

How is credibility verified?

Another signed-in builder must verify an outcome. Authors cannot verify their own answers, and each outcome can award points only once.

Which AI coding tools are supported?

All 41 Codeburn integrations are supported automatically, including Claude Code, Codex, Cursor, Copilot, Gemini CLI, OpenCode, Cline, Roo Code, Kiro, Devin, Goose, Warp, Zed, and more. Direct session logging remains available too.

YOUR MOVE

What are you
burning today?

Join the builders turning AI usage into a public sport—and useful work into real credibility.