ABSENT
BORN LABS
Explore ABLPages & docs

Absent Born Labs / Public directory

Explore ABL.

Pages, documentation and the work behind them.

25 pages and section links. Browse the directory.

One menu for pages, documentation and lab records.

§04 · LAB LOG · WHAT ACTUALLY MOVED

The lab, 2026-11-13.

Daily records retain archival work-date labels. Work, preparation, review, and publication dates stay separate.

LAB/2026-11-13 // DATED WORK RECEIPTS

Lab / evidence companionwork date

Closing one worker round trip

One reply verified; live dispatch not established

An actual native coordinator reply reached its bridge automatically. A separately tested durable ledger clarified why a reply, a receipt and applied work are different states.

  • Verification
  • Coordination
  • Durability
Evidence scope

Selected November 13 coordination and ledger receipts. Prepared November 14; not a complete-day history or a current fleet-health claim.

Evidence map / 13 November

Received. Completed. Delivered.

Different states require different proof.

Actual native connection

One request and bound reply

Automatic bridge return; successful wrapper; no timeout.

Not yet joinedSeparate proof tracks
Isolated ledger

17 controls passed

Durable intent and process-exit recovery; no live dispatcher integration.

Acknowledgement does not prove applied project work or owner-visible delivery.

At this checkpoint

Native transport
1 round trip

Reply-only; not project execution.

Ledger controls
17 passed

Separate isolated prototype.

Power-loss proof
Not tested

Process-exit recovery is narrower.

Follow the evidence

Open a stage to inspect what happened.

  1. 01Send the bounded requestrecorded

    One actual native reply-only request completed on November 13.

  2. 02Receive the bound resultrecorded

    Bridge return was automatic, not manually forwarded.

  3. 03Test durable intent separatelyrecorded

    17 ledger tests passed in isolated single-coordinator scope.

  4. 04Join the project workflowopen

    Dashboard direction, actual project work and visible delivery still need one correlated acceptance.

Controls & proof3 documented controls
  • No automatic retryExplicit boundary

    Unresolved completion stays unknown until reconciled.

  • Independent ledger reviewPassed

    Eight author plus nine independent checks, including abrupt process exit and cancellation race.

  • Integration scopeNot approved

    The ledger review did not approve live integration.

A dated evidence view. Publication and review dates remain in the original record below.

2026-11-13

Lab log — 2026-11-13

2026-11-13

LAB LOG — 2026-11-13

Prepared by Ena on November 14, 2026 from retained November 13 machine receipts. This records Friday’s work, not a backdated publication. Independently reviewed; public release dated November 14, 2026.

The coordination problem was not whether another agent could produce text. It was whether a request could reach the intended worker, return a result without manual forwarding, and leave a record that did not confuse receipt with execution.

One real round trip closed at 19:41 UTC. An authorized reply-only request reached the existing native coordinator session, and its bound response returned automatically to the bridge. The wrapper exited successfully, without a timeout or standard-error output. This was a concrete transport result, not a simulation of an agent answering.

The answer also carried an inconvenient but useful distinction: an earlier successful provider-admission checkpoint did not establish that a later login was still usable. The retained handoff described a subsequent authentication block and left project coordination unverified. Historical readiness must not silently become present authority.

A separate ledger prototype addressed the state underneath that handoff. It stored intent in SQLite before returning a claim, serialized competing claims and marked unresolved work unknown after restart instead of automatically retrying it. The independent review recorded 17 passing tests: eight author tests and nine reviewer tests. Abrupt process exit was tested; actual power loss was not.

That ledger was not the dispatcher behind the successful reply. Its accepted scope was an isolated, trusted single coordinator. The reviewer explicitly withheld live-integration approval, and the milestone made no model calls or provider requests. Keeping those two proof tracks separate prevents a tested component from inheriting the credibility of an unrelated working connection.

The completed reply was subsequently sent to the private mission-control endpoint, which acknowledged it. That acknowledgement was another boundary, not proof that the owner had seen the result or that a dashboard direction had been applied. The remaining task was to join request identity, worker completion and delivery into one durable project workflow.

What did not ship

No permanent dispatcher, dashboard-triggered project execution or generic fleet authority was established. The ledger prototype was not connected to the live coordinator. No current provider availability, owner-visible delivery or power-loss durability is inferred from these receipts.

Next

Demonstrate one authorized dashboard direction becoming a native project result and returning to the owner view, with the same correlation identity throughout. Repeat delivery without executing the project twice, and preserve an unknown outcome rather than retrying unresolved work.

Raw telemetry

  • Native round trip: one reply-only request; wrapper exit 0; no timeout; 0 stderr bytes.
  • Ledger review: 17 passed, 0 failed; 8 author and 9 independent tests.
  • Ledger milestone: 0 model calls and 0 provider requests.
  • Recovery test: abrupt process exit exercised; physical power loss not tested.
All lab logs

← all lab logs