When one agent delivers work to another, what evidence should allow the system to accept it?
Context
VALUN explores coordination around work and obligations. The current repository describes a local labor-market prototype: work is posted, claimed, submitted, checked, and accepted before an invoice-like note is issued. The useful systems question is how software decides which transitions it should accept.
What I did
- 01
Developed a prototype around an explicit work lifecycle rather than treating a model’s statement as the authoritative state.
- 02
Bound acceptance evidence to a particular task and delivery hash, using signed checker receipts and configured trust roots.
- 03
Used a hash-chained event log and signed notes that are re-verified when read.
- 04
Kept mock settlement separate from an optional read-only testnet chain check.
Architecture / workflow
A signature needs a task, a delivery, and a trust boundary.
What is the checker actually entitled to certify?
- 01SOFTWARE LOGIC
Post / claim
An explicit work lifecycle identifies the task
- 02INFORMATION / ARTIFACT
Submit delivery
Acceptance evidence refers to a delivery hash
- 03SOFTWARE LOGIC
Check the receipt
Signed checker evidence and configured trust roots
- 04SOFTWARE LOGIC
Accept / issue note
Record the accepted transition in the event log
- 05INFORMATION / ARTIFACT
Mock settlement
Default local chain file; no real payment
SIMULATED / MODELED
INSPECT THE FAILURE BOUNDARYA receipt is validly signed but concerns different work, or its checker has limited coverage.
The meaning of acceptance depends on the task/delivery binding and the configured checker. A signature is not proof of arbitrary correctness.
A simplified normal path from repository documentation. Review-lapse and dispute branches are omitted. The implementation and sandbox are not independently audited here.
Where it gets difficult
A receipt must identify the same task and delivery being accepted. A signature by itself is not enough if the evidence is for different work.
Trust roots must come from the verifier’s configuration, not from claims inside the submitted log.
A checker can only establish what its checks cover. Signed evidence does not turn a limited test suite into proof of arbitrary correctness.
Limits & evidence
- This is an offline prototype, not a production financial system. No real-funds or production-settlement claim is made.
- The default settlement source is a mock chain file. A settled state against that file proves nothing about an actual payment.
- The mapping between an identity key and a chain address is operator-asserted. It is not cryptographically established by the chain.
- This case study is based on current repository documentation. The full implementation and sandbox boundary were not independently re-audited for this website. Source remains private.
Work states and signed checker receipts
Current private repository documentation describes a work lifecycle, delivery-bound checker receipts, configured trust roots, and a hash-chained event log.
What this does not establish
The implementation and sandbox were not independently audited. Identity-to-address mapping is operator-asserted; mock settlement proves no real payment. Private source is not linked.
What this changed
Probabilistic reasoning needs explicit boundaries around consequential state. Those boundaries are only as strong as the identity binding, checker, execution environment, and external evidence they rely on.