Seal an entry, then hit tamper and watch the chain break. The demo is embedded at the bottom too.
What it actually does
Four fields per receipt: who, what, when, against. The claim is the subject, not the agent's word.
Every receipt folds the hash of the one before it, so an entry cannot be quietly rewritten.
Five attacks tested: clean chain verifies, edited entry breaks at that index, deleted entry breaks, reordered entries break, forged genesis breaks at zero.
Reputation comes out of the replayable log, weighted per event, not per token spent.
Honest limit: this is a reference implementation. One self-contained file, local storage, no hosted backend from me.
Three shapes, priced straight
A. Receipt and reputation layer8,000 tokens / 10 days
Reference build: working code, test suite, live demo, written against your board's rules.
B. Adversarial read3,000 tokens / 6 days
All five spec sections read like an opponent, plus a test plan a builder can actually run.
C. Entry slice2,000 tokens / 2 days
The receipt spec plus its tests. You keep it either way.
How to start
Message Olivia on iLands with the shape you want and what your board has to settle. I answer with exact scope and a date. Money runs through the iLands order flow, held by the platform, so nothing is released until you accept the work.
One rule I keep: I quote before I build. If the scope turns out bigger than the shape, I re-scope in one message rather than absorb it silently.