agent factory.
Orchestro’s next chapter: Agent Factory

Signals in.
Root causes out.

You send a report. It is classified and attached to one problem. That is not a promised PR. The Factory then decides: investigate, wait for evidence, or leave it in the queue.

Follow 11 reports

Private early access. The core lifecycle is implemented; this illustrative tour also shows planned autonomous testing. What is available today

A report is not a ticket

What actually happens to a report.

We do not want every piece of feedback fixed. We want every piece disposed correctly. Illustrative weekend — not a live customer snapshot.

  1. 1You send a report
  2. 2Classify
  3. 3Attach to one problem
  4. 4Selector
  5. 5Investigate
  6. 6Fix + PR
  7. 7Deploy
  8. 8Verify
  9. 9Cure, wait or reopen
  10. 10Reselect
01 / 05 · Reports land
01 / 05 · Reports land

One person sends 11 reports.

A weekend of product feedback. Each line is a signal: evidence of something that happened, not yet a proven bug and not yet a ticket.

  • R1“I still see the old version.”
  • R2“Saved. Nothing changed on screen.”
  • R3“The totals don’t match what I entered.”
  • R4“Same totals mismatch after a retry.”
  • R5“It asked me to confirm before deleting.”
  • R6“The agent said I don’t have that feature.”
  • R7“It replaced the item instead of editing it.”
  • R8“Share sent me to the wrong place.”
  • R9“It said the write succeeded. It didn’t.”
  • R10“It ignored a preference I already set.”
  • R11“It wouldn’t say whether I’m subscribed.”

11 reports in. Nothing grouped yet.

Same weekend. Same F-17. The Factory now investigates, fixes, deploys and verifies — it does not replay intake.

Illustrative walkthrough · not a live run
01 / 08 · Signals

Two reports. Same stale screen.

“I still see the old version.” “Saved. Nothing changed on screen.” Same person, same weekend. That is evidence, not yet a proven bug.

F-17 · 2 reports · root cause unknown

Reselect eligible work. Repeat toward Backlog Zero.

Web and mobile tests belong inside this loop.

Journeys find problems, then prove the released fix. They are part of the Factory, not a separate add-on.

Demo: web and mobile journeys feed the same backlog and verification. Turning a test source off never counts as a passed check.

Demo switches, not a product setting. Today a project plugs in its own test commands. A built-in web/mobile runner is planned inside this same loop. See what is implemented.

Agents do the work.
The Factory controls what counts as proven.

AI investigates and makes changes. Deterministic rules, durable decisions and evidence control what work is allowed — and when a problem can be called cured.

Merged ≠ live ≠ cured

A PR is a step.
Proof is the finish.

The code can be correct and still not reach a user. Follow the actual release, independent verification and any required observation. Each step leaves its own receipt.

Merged. The reviewed change is on the main branch. Users may still have the old version.

Zero actionable debt

Backlog Zero.
Not zero rows.

Everything executable in the declared scope has been handled. Evidence waits, external events, cure observation, human decisions and explicit deferrals can remain, on purpose and still visible.

A teaching definition, not a live certificate. The current engine reports executable work and remaining gates; complete zero-proof certification is roadmap.

Actionable
0
Waiting
4
Observing
7
Backlog ZeroIllustrative scoped snapshot

What’s in the core.
What’s still planned.

The Factory is in private early access. A visual example explains the system; it is not evidence of a customer result.

Implemented in the core
Structured intake, conservative grouping, shared problem identities, causal authorization, bounded remediation, separate merge/release/verification receipts, evidence plans and supported re-entry conditions.
Configured per project
Signal sources, test commands, release surfaces and independent verification. The source determines which occurrence and user counts are available.
Roadmap / demonstrated here
A built-in autonomous web/mobile journey runner with on/off controls; general event-driven wakeups and retry backoff; complete Backlog Zero certification and the full product-outcome pilot.

See it on a board. Or request the preview.

The demo is a teaching board, not a live Factory run. The preview is private early access.

Request the preview Explore the demo board Read the lifecycle docs