You are @you on GitHub. Your receipts go to Payments at Acme (the demo company).
- Write, done
- Sign, next
- Send
What did you get done this week?
Built checkout tests and ledger exports while debugging webhook performance
Built 2 things, figured out 1 problem and researched 1 topic. 5 AI sessions, $97.57 (Claude Code $85.89, Codex $11.68).
Written privately by 0gm-1.0-35b-a3b-sia on 0G, TEE verified, paid with the app's own 0G key.
Check it yourself
Each link opens the place where the record lives- 0G
The private AI that wrote this week ran in a hardware enclave
0G's TEE check of the providerOpens here
What you got
4 items, $97.57- BuiltCSV export for ledgersidea to code 23m · PR in 10m · merged in 33h$11.68
- BuiltE2e tests for checkoutidea to code 37m · code to commit 19m · PR in 12m · merged in 21h$77.78
- Figured outReason for slow webhook test$3.97
- ResearchedOptions for payouts compared$4.14
- Built $89.46
- Figured out $3.97
- Researched $4.14
Pull requests3 PRs, 3 merged. Always included when you send your receipt
- test: e2e for checkout$58.31acme-demo/payments-api#102, merged
- test: e2e for checkout$19.47acme-demo/web-app#101, merged
- feat: CSV export for ledgers$11.68acme-demo/infra#100, merged
Each one opens on github.com.
How this was writtenPrivate AI on 0G, 1 private AI call
Written by 0gm-1.0-35b-a3b-sia on the 0G Private Computer, in private mode: only providers running the model inside a TEE could answer, and every answer's TEE signature was checked. It was paid with the app's own 0G key.
The requests reach 0G over HTTPS through its router and are not end-to-end sealed today, so the router could read them in transit.
- Each session went in as a private digest: what you asked, what the AI said at the end, tool counts, the GitHub outcome and the cost.
- The AI answered with a category and a short title. Only those are kept, never the prompts.
- One more call saw only the titles and merged them into this list and its one sentence.
- 97616eeb-4a78-44dc-947c-f93faf7f2ff6
Where money leaked2 tips
- 60% of spend went to sessions that looped. 1 of 1 had runs of failing commands. After 3 identical failures, stop and re-plan, or fix the environment first.
- 97% of tokens were cache reads. 1 very long session(s) kept re-reading a huge context. Starting a fresh session per task keeps that cheap.
Cost by difficultyCompared with your team
Work is only compared with work of the same difficulty, never across people.
| Difficulty | PRs | You, per shipped change | Team median |
|---|---|---|---|
| 1 | $20.47 | $14.26 | |
| 2 | $44.99 | $25.43 |
How the week wentIdea to merged in 28h
Typical time between steps, from log and GitHub timestamps. No AI involved.
- 1. Idea to code30mover 2
- 2. Code to commit19mover 1
- 3. Commit to PR11mover 2
- 4. PR to merge27hover 2
From idea to merged: 28h typical, over 2. Time with the AI before writing code: 3h.
Pull requests
3 merged of 3 opened.
AI cost per shipped change (merged PR or push to main): $29.82
Where the money went
- Shipped $31.15
- Learning $8.11
- In progress $0.00
- Wasted $58.31
Waste, by reason
- Plans nobody acted on0
- Left on the AI's question1
- Went in circles1$58.31
- Code that never shipped0
- Answers stopped midway3
What the categories mean
- Built
- Made something new that works: a feature, page, tool or script.
- Fixed
- Fixed a bug or something broken.
- Figured out
- Understood or diagnosed something: how it works, why it fails.
- Researched
- Compared options, read docs or the web.
- Wrote
- Docs, posts, specs, emails or scripts in words.
- Planned
- Designed or planned the work.
- Wasted
- Went in circles or produced nothing usable.