Offer cells

An offer cell is a whole offer that can stand on its own numbers.

It is not a role. It is not a prompt. It is not a department name. It is the organism: one promise a specific buyer can buy, delivered by a path the cell can actually run, judged by a micro-P&L the rest of the market can see.

The same natural forces that run the portfolio run inside a cell. As above, so below. Bound, Sense, Act, Transform, Maintain, Defend, Select, and Continue do not become eight cell employees.

The seven parts

Every cell has to name these:

1. ICP — who this is for, narrow enough that a stranger can tell who is out.
2. Problem — the expensive situation that person already has.
3. Promise — what the cell will make true, in language the buyer would use.
4. Price — what they pay, and what that price has to cover.
5. Channel — how the cell reaches that buyer without borrowing someone else's list as a fantasy.
6. Delivery — the work that fulfills the promise, including the human gate.
7. Micro-P&L — whether the cell earns, burns, or waits. Capital reads this.

If one part is missing, the cell is a sketch. Sketches can sit in a notebook. They do not get capital.

Why the unit is a cell

A company that starts as a pile of agents has no object for the market to select. Someone has to ask "which offer is working?" and the answer comes back as a vibe.

A cell makes the question concrete. The market can keep a cell, starve a cell, or kill a cell. The founder can still override. The default is selection by verified customer commitment, not by how articulate the agent sounded this week.

That is also why a cell is end-to-end. Split ICP from delivery and you get a marketing particle arguing with an ops particle while the P&L belongs to nobody.

What a cell may do

A cell can draft, research, assemble evidence, and propose the next move inside its authority. It can ask the substrate for budget, identity, and the last working context. It can wait at a gate when the next action is send, spend, publish, or a production change.

Sense may observe without spending tries. Act may change the world only by paying. Transform decides what became fitness. Select reads that fitness. Packets sent are not fitness.

A cell cannot rewrite the constitution. It cannot spend past its budget. It cannot mint a new identity. It cannot treat a chat as proof of customer commitment.

What "verified customer commitment" means here

A like, a polite email, or a "sounds interesting" is not commitment. Commitment is a recorded act the cell can show: a paid invoice, a signed order, a deposit, a written yes that the founder has accepted as binding. The exact evidence types will be named as a commercial cell is designed. They are not named yet, because there is no commercial offer yet.

Until that evidence exists, the cell stays small.

How cells relate to each other

Cells can share the substrate. They do not share a blended P&L by default. If two cells need the same specialist, they request that work through founding topology: one owner per outcome, temporary specialists, independent verification, shared evidence. If a cell wants capital that another cell earned, Select reads both micro-P&Ls and the fitness evidence, then proposes. A cell owner containing its own cell does not impersonate portfolio Select. The founder still owns irreversible transfers.

The forces page covers the physics and the optional effectors. The loops page covers how often they get reviewed.

Current state

The first offer cell is a covering experiment: fractional sit-the-seat on posted full-time roles, running on try-unit budgets with a dated stop. Optional effectors sit inside Bound, Sense, Act, and Defend. Bound determines ICP and targets. Sense researches lists, enrich, and packets (sense_market, no try debit). Act spends tries from Sense-ready packets. The founder signs outbound. Packets sent are not fitness. The stance is experiment-until-fitness.

That covering cell is not a catalog item and not a commercial offer. Read what is not being built yet if you expected something to buy.