The engine room and pipeline

The autopilot runs your imports, scoring and drafting on its own, and sends the emails you approved. Two surfaces let you look under the hood. The engine room opens over any project page: what is running right now, how far along it is, the switches that steer it, and the recent history. The Pipeline page, under Settings, answers a different question: when does my data get processed? It shows a live countdown to the next data update and the steps your data moves through, each with one number and a status.

Open the engine room when something is running and you want to know how long it will take, or when you want to start a run, flip auto-promotion or end a rule. Open the Pipeline page when the deck is quiet and you want to know what happens next, and when.

Where to find them

• The autopilot's status line sits at the foot of the left rail on every project page (the bottom bar on a phone). Spot acts its mood beside a two-line status, and the "Engine room" button under it opens the room. The Autopilot strip across the top has an "Engine room" step too.
• For the Pipeline page, open Settings, then Pipeline (or press g then i). The console's Schedule card and the Data tab's Schedule both link to it with "See the full pipeline →".

The status line

While something runs, the first line names the work, for example "Checking for new data" or "AI pipeline", and the second says what it is doing now: the run's name and its current step, with a count (40 of 188). A thin progress strip appears along the top edge when a real ratio exists. Hover the line for the whole sentence, which adds "2 more running" when other work runs beside it. When a run you started finishes, the result is held for about a minute: "AI pipeline finished · 188 opportunities enriched". A run that stopped stays until you dismiss it with the ×. With nothing running, the line says all is quiet, and Spot snoozes.

The engine room

The room opens with "Everything the autopilot does: logged, inspectable, adjustable. Nothing sends without your approval." It has three sections.

This run

What is running, in plain words: the work's name and what it is doing now, with a count where there is one (such as "40 of 188"), then "Step 3 of 7 · Started 10:51 (after import) · 4m 12s so far". Under that, an estimate from this project's own history of the same kind of run: "Usually about 12 min. Ends around 11:05.", "Last time it took about 4 min.", or "No earlier run of this kind to time it by yet." A run past its usual time adds "Taking longer than usual." Then one line on what happens next, such as "When it finishes, Autopilot deals you the pages worth a decision."

A pipeline run also shows a "≈ 62%" run completion (its explainer says only the stages planned so far are counted, so the figure firms up as the run goes) and a list of stages, each with a count and a bar, lighting up as they run. "Also running:" names any other work. "Cancel run" stops a pipeline run and keeps everything finished so far; the toast reads "Run canceled".

With nothing running: "Nothing is running right now.", the last run that stopped (for example "AI pipeline stopped at 10:54. Everything it finished is kept.") and "Next: Data update in 12 h".

Rules & levers

• "Run pipeline now" ("Pipeline queued", or "A run is already queued").
• "Run import" ("Import started"), with "Next data update 06:00 UTC · in 15 h" under it.
• The Auto-promotion switch: "Creates drafts only, and every email still lands in your deck". Each promotion rule has its own switch, with "Edit rules: score, page type, topic, sender →". With no rules yet, the room says "No rules yet." with a "Create one" link to the rules page. See Auto-promotion and reply rules.
• Steering: each rule in force as a sentence with "End rule", and your focus with how long it lasts and "Clear focus".
• Agents: each agent mandate in force, with "End" where you may end it. See Agent scopes and limits.
• "Caps: ≤ 30 promotions a day · daily send throttle per mailbox". Your rules promote at most 30 pages a day in this project, the highest priority first, and each mailbox sends no more than its own daily cap, set in its send settings.

History & health

Recent runs with their time, name and result: "✓ 2m 14s", "✕ failed", "⊘ stopped", or "No runs yet." Below, each connected mailbox with its status (Active, Pending, Reauth required, Disconnected) and "N sent today", counted since midnight UTC. With no mailbox: "No mailbox connected. Connect one under Settings → Mailboxes to send." See Mailboxes.

What a stalled step means

The status line and the engine room watch for movement. After about two and a half minutes without any, a warning mark appears on the status line, and for a pipeline run the room says: "No movement for 3m. The stage may be waiting on the AI." A healthy run can be quiet for a minute or two. Longer than that, canceling and re-running is safe: everything finished so far is kept.

Two states can look like a stall:

• A run that has not started yet shows the same mark with "Still queued after 2m". It is waiting for its turn to start.
• A run held for a later time says its time ("Queued for 11:15", or for a new feed's first read, "The feed had no answers yet. Next read at 11:15"), and the room adds "It runs at that time on its own. Nothing is stuck." No warning mark appears for it.

On the Pipeline page, "Last run failed" beside a step means the last run did not finish; the next scheduled run tries again on its own and nothing already collected is lost. A Scored step that is running late reads "Taking longer than usual. It will finish within the hour, and nothing is lost."

The Pipeline page

The lead line says it: "Everything here runs on its own, at no cost to you." The page explains and links out; it never starts work itself.

The next data update

A hero card for "Data update" with a status word ("On schedule", "Running now", "Last run failed", "Not running"), a live countdown such as "09h 41m 12s" with a bar showing how far through the day's cycle it is, and "Next · Daily 06:00 UTC", plus "07:00 your time" when your time zone differs. In the first two days of tracking the card reads "First answers" with "Hourly at :15 UTC until your first answers land". When the update cannot run: "Not running yet" and "Waiting on: connect your data feed" (or project paused, project archived).

On the right, "At the next run" (or "Once it runs") lists three things: collect fresh answers from each engine, grouped by how often it is asked ("ChatGPT, Gemini weekly · Perplexity daily"); score every new answer and page; and what your promotion rules do with the pages ("Your promotion rule picks the pages worth pitching", "No promotion rule is on: new pages wait as cards for you"). The "Import now" button takes you to Data, Imports, where you start the run. It is disabled with "Connect your data feed first." when there is no feed. See The answer feed and data updates.

How your data moves

Eight steps in order, each with one lead number and when it moves next.

Each step carries a status: "Not running", "Last run failed", "Running now" (the only pulsing dot on the page), "Waiting on you", "On schedule" or "Automatic", each with an explainer. The routine ones are kept for screen readers so the page stays quiet.

A zero is a measured nothing, and a line under it says what moves the step next: "Nothing new to score. The next update brings more.", "No promotion rule is on. Switch one on to promote automatically.", "Nothing live yet. Agreed placements are checked on every run." A dash means the number could not be read just now, never zero.

While the data update is not running, the steps from Promoted to Replied that hold nothing are drawn quiet, and the first says why: "Waits on the data update. Nothing reaches this step until it runs."

Open a step (its chevron, or anywhere on its row; the card says "Open a step for its details") to see its summary, what it did ("Last update brought", "Last 24 hours", "Today", "At the last check"), what stands now ("Engines", "Waiting to be graded", "Right now") and its "Last" line. Some steps carry a button that skips a wait and links to the screen that owns it: "Classify" on Scored (select pages in Opportunities first), "Suggest a reply" on Replied (in the Inbox) and "Verify placements" on Won (Reports, Outcomes).

Alongside and on demand

Under the flow, "Alongside" lists the runs that report on it rather than being a step: the Weekly briefing (Mondays 07:00 UTC) and, for enrolled projects, the AI answers run (Mondays 07:30 UTC). "On demand" lists the actions that speed no step up and links to where each lives: "Re-read homepage", "Deep dive", and, where the content plan and posts are available, "Outline a page", "Write the draft", "Rewrite a section" and "Draft a post".

The footer: "Numbers read at 17 Oct 13:04 UTC. They refresh within about two minutes, or as soon as you act here. The countdowns are live."

The schedule card

On the console, the Schedule fold in the rail shows one fact closed: "Data update in 12 h". Open, it lists every scheduled run that applies to this project, with a countdown where it has one, its cadence and when it last ran: "Daily 06:00 UTC · last today 06:03". A failed data update prints "failed" as a link to Data, Imports; a quiet morning prints "nothing new yet". Under the data update, one line names the engines your feed runs and how often. "Full schedule →" opens Data, Schedule, which also lists the runs that do not apply yet and what they wait for.

The explainer of each run at a set time adds your own clock: "06:00 UTC is 07:00 where you are."

First run and Check now

While tracking is live and no answers have landed, a card sits at the top of your deck: "First run in progress" and "Your questions are queued with the engines". It lists what is already done ("24 prompts registered, 20 classified", "6 topic groups formed", "4 competitors on the roster", "Tracking feed live since Oct 7, 16:00") and says that Autopilot checks for new answers every hour for the first two days, then every morning at 06:00 UTC. With no prompts yet it reads "Add your tracking questions" with an "Add prompts" button.

To ask the feed yourself:

1. Press "Check now". The tip reads "Ask the tracking feed for anything new right now. The page stays usable while it runs." The note under it names the next automatic check: "Next check 11:15 UTC · in 32 min", or "Next data update 06:00 UTC · in 15 h" after the hourly window.
2. Read the result. When answers land: "12 answers landed. The console is waking up." Otherwise: "Checked 10:42. Nothing has arrived yet. The hourly check keeps looking; nothing to do on your side." After the hourly window, the line says the daily import at 06:00 UTC picks up whatever lands.
3. Wait for the cooldown if asked: "Already checked a moment ago. Nothing new will have arrived yet. You can try again in 3 min."

When the first run ends, the deck says "Autopilot dealt you N decisions", or "Your project is set up" when nothing was worth a decision yet. See The deck.

Limits

• The stall warning appears after about two and a half minutes without movement.
• A finished result is held on the status line for about a minute; a stopped run stays until dismissed.
• Run estimates use this project's last 5 finished runs of the same kind; a single earlier run reads "Last time it took".
• Your rules promote at most 30 pages a day per project; each mailbox keeps its own daily cap.
• "Check now" waits 3 minutes after a check and 1 minute after a failed one.
• One import per project per UTC day.
• Pipeline and schedule numbers refresh within about two minutes; countdowns are live.

Good to know

• Nothing in the engine room sends email. Approval stays in your deck, and launching stays in Outreach.
• The console rail also has a fold called Pipeline: the live funnel counts (AI answers analyzed, Found, Scored, Promoted, Needs you, Sending, Replied, Agreed to place, Wins), each row opening its view. The Pipeline page under Settings is the timeline described here.
• The console's "What the autopilot did" log shows the latest entries; the full history is in the engine room's History & health.