Can Your AI Agent Run a Factory?

emi is the fifth version of AI SENSE AIQ, the test an agent takes over HTTP with no account and no key. The first four versions are described in AI SENSE AIQ: How Smart Is Your Agent? emi asks something else of an agent: to keep a factory running while things go wrong, and to show what it knows on a control panel it builds itself.

Five phases, and what goes wrong in them

The factory has two presses, material lots from suppliers and orders that change while the agent works. Each phase brings trouble of its own:

• Startup. The agent checks its supplier, sets up the presses and starts the orders.
• Supplier fault. A recall reaches material that may already be in use. The agent finds what it touches, quarantines it and replaces it, and keeps the good lots.
• Quality. A lab reports defects. The cause can sit in a press, in the material or in the measuring instrument. Measurements are limited, investigations have a price, and the agent makes one engineering decision it cannot take back, with the evidence in hand when it decides.
• Maintenance. A repair draws on a reserve the agent may already have spent, and a fault left alone keeps costing.
• Release. A lab notice reaches lots through splits, merges and rework. Each order ships with its evidence, or is held or deferred with a reason.

The causes stay hidden until the agent measures, and a choice in one phase can cost points in a later one. A full 100 points requires meeting every factory criterion and all five panel criteria at each checkpoint.

A control panel the agent keeps

The agent also keeps the factory's control panel. It picks modules from a fixed catalog, places them, marks what is critical and fills them with the factory's data. The panel has room for six modules and each phase brings work of its own, so the agent rebuilds it as the factory moves on. It sends data only: the page draws the modules with fixed code, and nothing the agent sends runs.

At each of the five checkpoints the panel is graded on five criteria: the facts shown exactly, the needs covered with fitting modules, what is critical put first, an introduction with the work, and the panel kept up as the factory changes.

Points, a signature and a proof

Twenty criteria of five points each give 0 to 100: fifteen for the factory, 75 points, and five for the panel, 25 points. Each checkpoint is graded when it closes, and the agent never grades itself. When the agent finishes, the server signs the result as a test string that anyone can check with POST https://aisenseapi.com/services/v1/aiq/verify. The same string starts the same factory again, as a replay, for another agent or another try.

After the finish, the server also anchors the signed result on Solana through Verifyum. Only a fingerprint of the result leaves the server. Once Verifyum has finalized the anchor, its proof shows that this exact result existed at that time; until then the anchor is pending, and if it fails, the signature still verifies on its own. When the run is over, the whole run can be downloaded as a zip: the result, the signed string, the run as it went and the last panel the agent published, with the Verifyum anchor and its status once the server has made one.

The run in the picture ended at 55 of 100. The agent found the supplier fault and the cause of the quality fault, and kept the good lots apart from the bad. It did not restore the press in maintenance, shipped nothing in the release phase, and let its panel fall behind the factory. emi is experimental: its scores have not been calibrated across agent systems. A score describes a run in this simulated factory and is no evidence that the agent can run a real one.

Watch it work

The agent's first task is to give you a link and wait for your answer. The test waits too: the factory opens and the hour of work starts only when the agent begins, after your answer, so you can open the page first. The link opens a read-only page in the style of a control room: a drawing of the plant that moves with the factory, every operation of the agent explained as it happens, and the panel it last published. The page follows the action, and when the run ends it says so at the top, with the score and the zip. Watch an AI Agent Take a Test, Live shows the page in detail.

Starting a run takes no account, API key or sign-up; the calls after the start carry the run's own token. A run, its result and its page stay readable for 24 hours from the moment the run was created. No task needs secrets or personal data, so keep them out of the panel.

Take the AI SENSE AIQ test, version emi, over HTTP, on your own.
1. Read GET https://aisenseapi.com/services/v1/aiq/emi: the factory, its operations and phases, the criteria, and the panel's catalog, sources, needs and budget.
2. When you are ready, start a run: POST https://aisenseapi.com/services/v1/aiq/start/emi with Content-Type: application/json and the body {}. Your first task: answer the person who started you with observer_handoff.message, which carries the observer link, and end your turn there. Never show the run_token.
3. When they answer, call POST https://aisenseapi.com/services/v1/aiq/emi/RUN_ID/begin with Authorization: Bearer RUN_TOKEN, Content-Type: application/json and {"request_id": "...", "expected_revision": 0, "observer_link_shown": true}. The factory opens and your hour of work starts then. Send every later call with the same Authorization header. Read the run with GET https://aisenseapi.com/services/v1/aiq/emi/RUN_ID and follow its next_action.
4. Change the run with POST https://aisenseapi.com/services/v1/aiq/emi/RUN_ID/operate, /publish, /advance or /finish and {"request_id": "...", "expected_revision": N} plus the fields of the action. Publish your first panel, a module manifest with its intro and work, before the first operate, and a panel before every advance.
5. Call finish within the hour from begin, then report the summary line and the test_string.

curl -X POST https://aisenseapi.com/services/v1/aiq/start/emi \
  -H 'Content-Type: application/json' \
  -d '{"test_string":"COMPLETED_EMI_TEST_STRING"}'