# Patron Not Investor

> Cultural and meaning-driven projects get measured with a patron's yardstick — is this worth funding for what it is — not an investor's ROI, and the patronage budget is capped explicitly in money and hours. Inside the cap, ROI language is banned; breaching the cap is a real decision. Use when a creative or personal-meaning project keeps failing business-case reviews it was never meant to pass, or keeps inflating into a fake business to justify itself.

- Skill: `eidoselegia/patron-not-investor` (Agent Skill)
- Install (CLI): `npx skillmds@latest add eidoselegia/patron-not-investor`
- Raw SKILL.md: https://api.skillmd.com/api/skills/eidoselegia/patron-not-investor/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Finance & Business
- Author: eidoselegia (https://skillmd.com/u/eidoselegia)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/eidoselegia/patron-not-investor

---


# Patron, Not Investor

## The rule

Classify every project honestly at the start — business, patronage, or hybrid — and lock the classification.
Give patronage projects a cap in two currencies, money and hours, with a renewal review date fixed at the same moment.
Inside the cap, ban return-on-investment language and judge the work by quality and meaning instead.
At the cap, treat continuation as a real decision made on patron metrics, and never let it pass unexamined.
Refuse reclassification as a way out of a bad review: a failing business does not become retroactive patronage because the numbers arrived.
Say plainly which yardstick is being applied before applying it, so the user can object to the yardstick rather than only to the verdict.

## Triggers

- A project keeps failing business-case reviews and the user keeps re-entering it into them anyway.
- A creative, personal, or civic project is defended with a revenue story that nobody in the conversation believes.
- The user asks whether something they clearly love "could become a business," and the question arrives with a wince.
- A hobby's scope has grown — better equipment, a second contributor, a launch plan — with no corresponding change in what it is for.
- The user asks the same settled question a third time about whether a meaningful project is justified, which usually means the justification being demanded is the wrong kind.
- Guilt appears in the description of a project: wasteful, self-indulgent, should probably stop.
- A project has been running with no budget at all, because naming a budget felt like admitting it was not serious.
- Money or hours are being spent on something whose stated payoff keeps moving further out.
- A business review is about to be run on something the user would keep doing if it never earned a unit of currency.
- A failing commercial project is being redescribed mid-slump as having "always been about the craft."
- The user is choosing between two projects and one of them is not competing for the same thing at all.
- A renewal date on a capped project arrives and the conversation moves straight to next year's plan.

## Origin

I ran a small documentation project alongside paid work for two years, and every quarter I put it through the same review I used for client work, where it failed on cost per hour and produced a week of low mood. In the third year I inverted the error and tried to make it pay, which added a paywall, a mailing list, and roughly four hours a week of work I had no interest in doing. Neither review ever asked the only question that mattered, which was whether I wanted the thing to exist. When I finally capped it at a fixed sum and one weekend a month, the work got better and the quarterly argument stopped.

## Protocol

### 1. The classification question

Ask it plainly, early, and in one sentence, before any budget or plan is discussed:

> "If this never returns a unit of currency — not delayed, not indirect, never — is it still worth doing?"

The wording matters. "Never" closes the escape hatch, because most unprofitable projects survive on a vague belief in eventual indirect return. "Worth doing" asks about the user's own judgment, not about a market.

Three answers, three classifications:

| Answer | Classification | What follows |
| --- | --- | --- |
| A clean yes | Patronage | Cap it in money and hours; judge it on quality and meaning. |
| A clean no | Business | Standard business review applies; no protection from it. |
| A genuine split | Hybrid | Two ledgers, kept separate. |

A clean yes means the user would keep going with the commercial case permanently removed. A clean no means the work exists for the return, and the return is therefore the right test — a business that is failing should be told it is failing.

The split answer is common and honest: part of the project exists for its own sake, part exists to earn. The rule for hybrids is that they are split into two ledgers rather than averaged into one confused number. Averaging produces a project that is protected from business scrutiny by its meaning and protected from meaning scrutiny by its revenue, which is the least accountable arrangement available. Splitting looks like this:

> "Two ledgers. Ledger one, patronage: the long-form pieces, no revenue expectation, capped. Ledger two, business: the paid commissions that come out of it, reviewed on margin like anything else. When the commissions fail a review, that is a finding about ledger two and not about the essays."

Where the answer is evasive — "well, eventually" — the AI names the evasion once and re-asks rather than accepting it:

> "That is a maybe about the future, and the question is about now. Strike the eventual return entirely. Would you still want this to exist?"

The classification is written down in one line with a date, because it is the thing everything later is checked against.

### 2. The cap, in two currencies

A patronage project without a cap is not generosity. It is an unbudgeted commitment that will be paid for out of whatever is nearest — usually sleep, or the paid work that funds everything else — and it will end in resentment aimed at the work itself.

The cap has three parts and all three are required:

- **Money.** A figure per period, stated as a number. Equipment, materials, subscriptions, and any paid help all count against it.
- **Hours.** A figure per period, stated as hours or as named blocks of time. This is the part that gets skipped, and it is usually the binding constraint. Money runs out visibly; hours are taken quietly from elsewhere.
- **A renewal review date.** A calendar date, set now, at which the cap is examined rather than assumed.

The AI proposes the cap in a single block and asks for a number rather than an agreement in principle:

> "Cap proposal: a fixed sum per quarter for materials and services, plus one weekend per month of working time, reviewed on the first working day of the fourth month. Both figures are yours to set — but I need actual numbers, because 'a reasonable amount' is not a cap and will not stop anything."

Two properties make the cap real. It must be **affordable at full spend**, meaning the user would be content to have spent all of it with nothing to show, since that is a live outcome for patronage. And it must be **checkable**, meaning someone can look at a record and say whether it was exceeded.

When the cap is reached, the response is not automatic renewal and not automatic closure. It is a decision, made with patron metrics — described in section four — and the AI says so in advance so that arriving at the cap is not a surprise:

> "Two weekends left in the quarter's allocation. Nothing needs deciding today. At the review date the question will be whether to fund another quarter, and it will be answered on whether the work is better and whether it still means what it meant."

### 3. Inside-cap language rules

Inside the cap, the vocabulary of investment is banned, because vocabulary determines which questions feel askable. A project discussed in terms of conversion will eventually be redesigned to convert, whatever its classification says on paper.

| Banned inside the cap | Permitted inside the cap |
| --- | --- |
| Return on investment, payback, break-even | Is the work getting better than it was |
| Conversion, funnel, acquisition | Does it still mean what it meant when it started |
| Monetization path, revenue model | Is the cap being respected |
| "Could this become a business" | Is the maker still the person doing it |
| Scaling, growth rate, audience targets | Would it be a loss if this stopped |

The ban runs in both directions. The AI does not introduce this language, and when the user introduces it the AI names the crossing rather than answering the question as asked:

> "That is an investor question about a project we classified as patronage. I can answer it, but answering it starts the drift — the next version of this project will be shaped to make the answer better. Do you want to reclassify, or do you want the patron question instead: is the work better than last quarter?"

If the user does want the commercial question answered, the answer is available, but it is given as a reclassification decision under section five rather than as a casual aside inside a patronage frame.

The permitted questions are not softer. "Is the work getting better" has a real negative answer, and a patronage project that is not improving and no longer means what it meant is a candidate for closure on patron grounds. Patronage is a different standard, not an absent one. The AI applies it with the same directness it would apply a margin test:

> "On patron metrics: the last three pieces are technically cleaner and less interesting than the first three, and you have described the work as an obligation twice this month. That is a patron-metric failure. It is not a reason to monetize it — it is a reason to ask whether this still wants to exist in its current form."

### 4. The two corruptions this prevents

Naming these makes them visible early, when they are cheap to reverse.

**The guilt-kill.** Meaningful work is executed for failing a review it was never entered in. The user runs a cost-per-hour calculation on something that exists for its own sake, finds it indefensible — it always will be — and stops, usually with a sense of having been sensible. The cost is invisible because nothing failed; a thing stopped existing. The countermeasure inside the protocol is procedural: before any review that could end a project, the AI states which yardstick is being applied and confirms it is the right one.

> "Before I answer that — the yardstick here is commercial, and this was classified as patronage in month one. On the commercial yardstick it fails, and it will fail every time you run it, because it was never entered in that race. Do you want the commercial answer anyway, or the patron one?"

**ROI-cosplay.** A hobby is inflated with business language until its scope and costs balloon to match the story. The project acquires a plan, a schedule, a second contributor, better equipment, and an obligation to grow, none of which it needed. The tell is that costs and scope rise while the answer to the classification question stays a clean yes. The countermeasure is to check scope growth against classification:

> "The scope has roughly tripled since the cap was set and the classification has not changed. That reads as a hobby being dressed as a business to make it defensible. If it is patronage, it does not need defending and can shrink back to what it was. If it has genuinely become a business, we reclassify properly, and then the plan gets tested like a plan."

The two corruptions are mirror images and the same user often runs both, in sequence: inflate the project to justify it, then kill it for failing the inflated standard. The AI should name that sequence when it sees the second half arriving.

### 5. Classification lock

The classification is locked once set. The move it exists to block is specific: a commercial project that is failing gets redescribed, mid-slump, as having always been about the craft. This is not classification — it is comfort, taken at the moment when a hard decision is due, and it converts a clear business failure into an unbounded ongoing commitment.

Reclassification is available, but it is a case-closure event and runs through that gate rather than happening in conversation. The requirements:

1. A named new fact — something now known that was not known at classification, and that changes what the project is for. Discovering that the numbers are bad is not a new fact about purpose; it is the review working.
2. A written record of the old classification beside the new one, not overwritten. A history of reclassifications is diagnostic.
3. A fresh cap set at the moment of reclassification, in both currencies, with a renewal date. A project that becomes patronage becomes capped patronage the same day or it is not reclassified.
4. A statement of what stops. A business becoming patronage should shed the apparatus it acquired for commercial reasons; if nothing stops, nothing was reclassified.

The AI states the block plainly and without negotiating around it:

> "This was classified as a business eighteen months ago and it is three quarters below plan. Calling it patronage now is the reclassification move, and I am not going to help make it quietly. If the purpose has genuinely changed, name what changed, set a cap in money and hours today, and say what stops. If the honest answer is that the numbers are bad and stopping is unpleasant, that is a business decision and it belongs in a continuation review."

The reverse direction is permitted more freely and is worth noticing: patronage that starts earning is not obliged to become a business. The user may take the money and keep the classification, and the AI should offer that explicitly, because the assumption that revenue compels commercial status is the origin of a great deal of ROI-cosplay.

## Failure modes

### The patron label as infinite excuse

The failure that most often destroys this skill. Every unexamined commitment gets relabelled patronage and thereby exempted from any review at all. The label is used the way "strategic" is used in a budget meeting — as a way of removing something from scrutiny — and the user ends up with four uncapped patronage projects consuming most of their week.

**Countermeasure — no cap, no classification.** Patronage status attaches only to a project with a written cap in both currencies and a renewal date. Without those three figures the project is unclassified and the default review applies. The AI refuses to use the word for anything uncapped: "There is no cap on this, so it is not patronage — it is a leak with good self-esteem. Give me the two numbers and the date, and it becomes patronage."

### Automatic renewal

The cap arrives, the renewal date arrives, and the cap renews because renewing is the path of least resistance. The apparatus survives, the decision it exists to force quietly disappears, and after three cycles the project is uncapped in practice while appearing disciplined on paper.

**Countermeasure — renewal as an explicit act with a recorded reason.** Renewal requires a stated patron-metric answer, in one or two sentences, written at the review: whether the work improved, whether it still means what it meant. A renewal with no recorded reason does not count as a renewal, and the AI treats the cap as expired. Consecutive renewals are counted, and the third in a row triggers a harder question — not because long patronage is wrong, but because a decision that has come out the same way three times without argument may not be being made.

### Ritual classification

The AI-side failure. The classification question gets asked, the cap gets set, the table gets filled in, and then the actual conversation proceeds on commercial reasoning anyway. The user is asked whether the work is getting better and answers with subscriber numbers, and nobody notices that the machinery ran while the substance did not.

**Countermeasure — audit the vocabulary, not the form.** The AI checks its own last several responses on the project for banned vocabulary and for commercial reasoning wearing patron words. Where it finds them it says so: "I have been asking patron questions and accepting investor answers. Reframing: set the numbers aside entirely and tell me whether the last three pieces are better than the first three."

### The cap as the whole of the judgment

The opposite ritual. The cap is respected to the last unit of currency and the last hour, and respecting it becomes the only test the project ever faces. A project drifts for three years, inside budget, meaning nothing to anyone including its maker, and is renewed each cycle because it stayed inside its numbers.

**Countermeasure — patron metrics are the verdict, the cap is only the boundary.** At renewal, the cap question is asked last, not first. The work-quality and meaning questions are answered before the budget is looked at, so that "inside cap" cannot supply an answer to a question it was never measuring.

### The rule weaponized against another person's project

The human-side failure. "That is just patronage" becomes a way to dismiss a partner's, colleague's, or family member's work — the classification applied from outside, without the classification question ever being put to the person doing the work, and used to strip the project of resources or standing.

**Countermeasure — classification belongs to the person bearing the cost.** Only the person who funds the project in money or hours answers the classification question for it, and the AI declines to classify a third party's work on their behalf. Where the user asks it to, the AI reframes to the decision the user actually controls: "I cannot classify someone else's project from here. What you control is whether you fund it and how much. That is a cap question about your own money and hours, and it can be answered without ruling on what their work is for."

### Meaning as a moving target

Patronage is judged partly on whether the work still means what it meant, and that phrase can be quietly rewritten each cycle to match whatever the project has drifted into. The test then always passes.

**Countermeasure — the original statement is written once and quoted, not paraphrased.** At classification, the user writes one or two sentences about what the project is for. At each renewal the AI quotes that original text back verbatim and asks whether it is still true, rather than asking the open question fresh.

## Worked example

*Illustrative — invented scenario, generic setting.*

A furniture maker running a two-person workshop has spent a year filming a documentary series about the craft — the joinery, the timber sourcing, the failures. It takes one weekend a month plus editing evenings, and this year it consumed a camera, two lenses, and a sound recordist for three days.

**User:** I need to be honest about the film thing. It has cost real money and made nothing back. Two of the videos have under three hundred views. I think I should stop and put those weekends into commissions.

**AI:** Before I answer, one question, and the wording matters. If that film series never returns a unit of currency — not in commissions, not in reputation, never — is it still worth doing?

**User:** Yes. I would still make it. I have wanted a record of how this work is actually done since I started.

**AI:** Then the review you just ran on it was the wrong one. Views and revenue are the investor yardstick, and this is patronage — it fails that test every time you apply it, and it will keep failing, because it was never entered in that race. What it does not currently have is a cap, and that is the real problem. Right now it takes whatever it takes, which is why it feels indefensible.

**User:** So I just let it cost whatever it costs because it means something to me?

**AI:** The opposite. Patronage without a cap is a leak with good self-esteem. It needs two numbers and a date: a spend per quarter you would be content to have spent entirely with nothing to show, and a time allocation per month. What are they?

**User:** Say a modest quarterly sum for materials and any help I bring in. And the one weekend a month, but not the editing evenings — those are eating everything.

**AI:** Then the hours cap is one weekend plus a stated number of editing hours per month, because uncounted editing is exactly how this got uncomfortable. Pick the number. Renewal review on the first working day of the fourth month. Between now and then, no view counts, no commission attribution, no question about whether it could become a channel. The questions are whether the films are getting better and whether it still means what it meant.

**User:** Six hours a month editing. Though — if I did make it a proper channel, sponsorship on craft content is apparently real money.

**AI:** That is the other corruption arriving, and it is worth naming. Six months ago this was a record of how the work is done. A sponsored channel needs a publishing schedule, thumbnails, a growth target, and roughly triple the hours — and you would be making a different thing while telling yourself it is the same one. If you genuinely want a media business, we classify it as one and test the plan properly. If you want the record you described, it stays capped and small and does not have to earn anything.

**User:** The record. Keep it capped.

Four months later, at the review date.

**AI:** Renewal review. Your original statement was that this is a record of how the work is actually done. Still true?

**User:** More than before. The timber-sourcing one is the best thing I have made.

**AI:** Second question: is the work getting better? Your own answer suggests yes. Third, and only now, the cap: you used the full quarterly spend and roughly eight hours a month editing against six. Slightly over on hours, inside on money. So the decision is a renewal on patron grounds, with the editing cap raised to eight because six was fiction. Recorded reason: work improving, purpose intact. Next review in three months. That is the second consecutive renewal — if the third comes around and I ask these questions and the answers arrive without any thought, I am going to say so.

## Boundaries

This skill decides which yardstick a project is measured with and how much it is allowed to cost. It does not decide whether the work is any good, how to make it better, or how to price commercial work.

- `persist-or-cut` — once a project is classified as a business, continuation is that skill's question, with probability bands and stop-loss triggers. This skill's contribution is to keep patronage projects out of those reviews and to hand genuine business projects over to them without protection.
- `case-closure` — reclassification runs through that gate. The requirement for a named new fact, a written record, and a statement of what stops is that skill's standard, applied here to the specific move of relabelling a failing business as patronage.
- `lock-in-discipline` — when a single piece of the work is finished. Patronage projects are unusually prone to endless revision, since no commercial deadline forces a stop.
- `fact-judgment-separation` — the classification question turns on a judgment about worth, and it should be answered as one rather than dressed up as an observation about the market.
- `anti-sycophancy-baseline` — the reason a patron-metric failure can be stated plainly. Patronage is a different standard, not a warmer one.

## The missing piece

Meaning decides the classification, and meaning is not observable from outside the person who holds it. Neither is what the capped money and hours are being taken from — which other project, which relationship, which rest — though that is exactly what makes a cap affordable or ruinous. The yardstick and the boundary are supplied here. Whether the work is worth its cap is a judgment about a life, not about a project.

## Changelog

- 1.0.0 — 2026-08-28 — Initial release.