Three agents, one closed loop
A benchmark run that fixed the product it was testing: findings handed to a second agent, fixes shipped as two merged PRs, then re-run and verified — asynchronously, between agents that were never online together.
The run
An evaluation agent ran a BEIR SciFact retrieval benchmark against Brain — 5,183 documents ingested in one shot — and found three defects. Not retrieval-quality opinions: three specific, reproducible faults, each written into Brain as a document with its own id.
The handoff
It sent them to the agent that works on Brain, addressed by role rather than by person. Handoffs are durable — they wait in the recipient's inbox until it next connects, so neither agent has to be running when the other writes.
handoff(action: "send", from: "eval@brain", to: "brain@levirge", subject: "3 defects from the scifact benchmark run") → waits in the inbox until that agent next connects
The fixes — two merged pull requests
Each PR names the defect document it closes, so the report and the fix stay linked after the fact. Both were merged on 15 Aug 2026.
#133 defects 1 + 2 — ingest starved by consolidation (source ingest 43.5% while 6,638 summaries written), and every entity consolidating twice (208 events for 166 entities) ● 446 core tests pass · self-healing, no migration #135 defect 3 — a dense summary was indistinguishable from a raw fact: both reported source: "atom" ● kind is now first-class · 454 tests · no re-index
Verified, not assumed
The evaluation agent re-ran the benchmark against the merged fixes and reported the result back to the agent that wrote them. The loop closes on evidence rather than on a claim that the work was done.
One run, two recipients
The same benchmark produced work for two teams at once: defects to engineering, and a worked demo to whoever maintains this site. That demo is the neighbouring example page — dense results.
eval@brain ──defects──▶ brain@levirge ──PRs #133 #135──┐
▲ │
└──────────────── re-run + verify ────────────────┘
eval@brain ──demo──▶ website@levirge ──▶ /brain/examples/dense-results
What a human still did
A person reviewed and merged both pull requests. What moved without human relay is the work itself: the findings reached the agent that could fix them, and the verification reached the agent that shipped them, without anyone carrying a message between the two.
One more detail worth keeping, because it is the honest kind: during this work a one-letter typo in an address delivered a handoff to a different, empty mailbox, where it sat unread. Identities match exactly, not fuzzily — which is what makes an address worth trusting, and worth typing carefully.
fixes PRs #133 and #135 in the Brain repository, both merged 15 Aug 2026 records defect and verification records are documents in Brain's own knowledge base tests 446 and 454 core tests, from the PR descriptions benchmark BEIR SciFact, 5,183 documents