The incubator desk
Six stations, one continuous bench
Everything AI-INTERF LLC does happens at the Saint Louis incubator desk at 4902 Moccasin Ter, Saint Louis - 63123-5618, United States (US). The desk is small, the trays are limited and the work is written down. Below are the six stations of the bench, each described in the order a team usually meets it, followed by the way a full placement moves from intake to a public demo.
Station APitch Sheet Bays
A pitch sheet bay is a narrow slot on the desk rail that holds one live sheet per placement. The sheet is not a brochure and it is not a business plan. It carries three things on a single page: the integration problem in plain language, the claim the team believes it can prove, and the smallest demonstration that would make the claim visible. Every team writes the first sheet by hand at the bench, and every later version is dated and kept.
The reason for a bay is discipline. Early systems design tends to sprawl, and a sprawled argument cannot be tested. By forcing the whole placement onto one sheet, the bay makes the weak link obvious. If the claim needs four paragraphs of setup, the claim is not ready. If the demonstration needs a friendly operator, the demonstration is not a demonstration. Rewriting the sheet each round is how a team learns to say the same true thing in fewer words.
Bays are also the place where a team learns to separate the integration story from the product story. Reviewers rarely fund a diagram, but they will remember a clean seam between two systems that used to hate each other. The bay keeps that seam front and centre and pushes marketing language to the margin.
Station IMentor Note Runs
A mentor note run is the written residue of a sitting. After a team meets a mentor at the desk, the mentor walks a fixed run of notes: what the team asked, what was suggested, what was rejected and why, and what the team said it would do next. The run is short, usually under a page, and it is filed the same day. Teams keep their own copies and the desk keeps a working set.
The fixed run matters because memory is a poor archivist. An undocumented suggestion becomes a rumour, and a rumour becomes a plan nobody agreed to. When a milestone review arrives, the note run is the only honest record of how the team got where it is. It also protects the mentor as much as the team: a written note makes clear what was advice and what was never promised.
Note runs are not performance reviews and they are not scores. They are the low, boring, essential layer of a healthy incubator, the equivalent of a lab notebook. Teams that keep their runs well tend to move faster, because they stop relitigating the same decision every two weeks.
Station FDemo Plan Desks
A demo plan desk is where a demonstration is designed before it is built. We start from the last ninety seconds that a reviewer will actually watch, and we work backwards from that moment. What must be on the screen, what must be true in the data, what happens if the network drops, and what the team says when something fails. The plan is written as a checklist with a fallback path attached to every step.
Most failed demos fail in planning, not in engineering. A team rehearses the happy path twenty times and never rehearses the awkward question. The desk insists on a failure script, a short one, spoken out loud, so that a wobble does not become a silence. In our experience a team that can narrate its own failure looks far more competent than a team that has only seen success.
The desk also keeps demo plans short. A plan that needs three machines, two colleagues and a live partner system is fragile in ways nobody can control. We prefer a plan that runs on one machine, with recorded data, and a single honest seam shown live. That is what people remember.
Station SOffice Hour Windows
An office hour window is a short, bookable slot at the Saint Louis desk, usually forty minutes, open to cohort teams and to outside teams when the bench has room. It exists for the question that does not fit a scheduled meeting: the data model that keeps drifting, the contract clause that keeps being postponed, the partner who keeps changing scope. One mentor, one team, one question.
Windows are deliberately not sales conversations. Nothing is pitched and nothing is upsold. If the honest answer is that the team should stop and rebuild a layer, that is the answer we give. Outside teams often use a single window to sanity check an architecture before committing a quarter to it, and that is a good use of the bench.
Booking is simple. A team writes to the desk with the question in two sentences and a preferred week. We answer with a time or with a note that the question belongs elsewhere. We keep the windows limited so that the mentor arrives prepared.
Station PMilestone Review Rounds
A milestone review round happens every few weeks, and it has a fixed shape. The team presents the current pitch sheet, the desk reads the accumulated note runs, and together we compare the claim to what the tray can actually show. A review round ends with a written decision: keep growing, cut a branch, or return a section to the soil and restart it smaller.
The word round is important. Reviews are not pass or fail gates. A round is a season, and a season can be honest about a thin crop. Teams that treat reviews as exams tend to hide problems until the end, and hidden problems are the expensive ones. We would rather see a team say plainly that a seam did not hold than watch a polished demo cover a fragile core.
The written decision is filed with the note runs and travels with the team. If the work later moves to another mentor, another firm or another city, the decision trail goes with it. Credit for the work belongs to the team, and the desk holds itself to the same standard of record keeping that it asks of the team.
Station DCohort Sittings
A cohort sitting is the moment the trays are moved to the long bench and six teams sit together. Each team gives one short demo and hears the others. There is no scoring and no ranking. The sitting exists so that a private prototype meets outside eyes before it meets a stranger with a cheque, and so that teams learn to ask each other the questions reviewers will ask later.
Sittings also create a small, honest community. Founders in the same season share vendor problems, hiring worries and integration pain that no single team can solve alone. The desk simply provides the bench, the time and a written summary of the response each team received, so that feedback is not lost when the room empties.
We keep the cohort at six teams. A larger table turns a sitting into a conference, and a conference is where honest feedback goes to die. Six is the number at which a room can still disagree out loud.
Process overview
A placement from walk in to walk out
The six stations are not sold separately. A team that joins the desk moves through all of them in a sequence that usually spans one to two seasons, and every step leaves a document behind.
Bench intake
We draw the integration surface as it actually is, including the parts nobody wants to talk about.
First tray
The first pitch sheet is written and the first demo plan is sketched at the desk on the same day.
Weekly rounds
Office hours, note runs and small demos repeat until the sheet and the tray agree with each other.
Review round
A review round cuts what cannot grow and names what should be replanted in a smaller shape.
Cohort sitting
The team demos at the long bench and receives a written summary of the room response.
Ask the desk a direct question
If you are unsure which station you need, write to us with the problem instead of the service name. We will tell you honestly whether the bench is the right place for it, and we will say so quickly.