Market sizing

Rank where to enter → count the units it covers → price them → check against a named top-down figure → write .claude/context/market.md. Output is one bottom-up number you can defend factor by factor and a ranked list of entry wedges. Asked for both flows, run wedges first, then size the top two or three, plus idea.md's wedge if it isn't among them: the slide is sized on that one.

What investors and filed decks ask of a market slide: method-sources.md. Traps: traps.md. Read it before adding anything up.

Context handoff

• At start, read whichever of these exist, and only ask what they don't answer:
  • .claude/context/market.md: a re-size. Re-open every source before reusing its number.
  • idea.md (format): the founder's wedge, a candidate in the ranking.
  • positioning.md (format): the buyer and the alternatives.
  • pricing.md (format): the price and the floor cost.
  • build.md (format): What works today sets each wedge's distance D; Traction is customer evidence (Kamps: a valid alternative to a formula).
• The price ceiling is what the alternative costs the buyer. No context file stores it: work it out with pricing's cost of the bad alternative, or ask.
• On completion, write .claude/context/market.md (format) for pitch-deck's market slide and fundraising's diligence answers. The slide is sized on idea.md's wedge, the one the deck pitches; if the top-ranked wedge differs, say so under Open decisions and leave idea.md to idea-sprint.

Rules for both flows

• Units before dollars. Count accounts and units from a dated source, then price them.
• Only what was opened counts. A snippet-only number is secondary: never summed, never on a slide.
• Every number is one line in the number-with-source format: [inferred] for a reading, (est.) for a calculation written out on the same line.
• Pages for other people hold principles and the result. Process stays in the notes.

Flow A: sizing

1. Name what the market is: the spend the product replaces, not an industry category. Ceiling = what the alternative costs the buyer; floor = the marginal cost from pricing.md.
2. Bottom-up first. Accounts × units per account per year × price; a monthly count is converted on its own line, never with a loose × 12. With a bps price, the unit is the volume per account per year. The beachhead is the annual revenue at 100% share of the first market (Aulet).
3. Top-down only checks, inside market.md and never on the slide: a named firm, its latest edition, the same cut, within an order of magnitude of the bottom-up.
4. Three kinds of money, never added together (money-types.md): C, current spend on the check or work the product would do; S, spend on the activity it touches; R, money at risk, which goes on the problem slide. TVL, volume and market cap are Context, outside every sum.
5. Bands. Beachhead US$20M to 100M a year (Aulet's 2012 draft, marked preliminary); vision above US$1B with the most persuasive evidence (YC). Outside the band, run the reverse calculation and label it calibration, never a price.
6. SOM = reachable accounts × units per account per year × price, never a share of TAM. Expansion from a material improvement (Gurley) is a mechanism with no number.
7. Declare what is weak: source concentration, data years, the token or FX price and its date, cuts with no number, and the billing unit.
8. Propose the slide the way pitch-deck's Market row asks: one bottom-up number, the multiplication with each factor sourced, the source on each card. Why now is its own slide. The price and the number are the founder's call.

Flow B: wedges

1. Vision, cycle and pillars, one line each (qxa-rubric.md).
2. Filters before the search, from how the customer operates: their decision depends on what the product observes or does; the gap costs more as they grow; solving it builds the vision. One real worked case each; the founder approves them before any search.
3. Five blind fronts in parallel, one by analogy, searches capped, reports kept whole: search-front-prompt.md.
4. Merge and filter before listing, then one card per candidate.
5. Rank by Q×A, D only breaking ties (rubric).
6. One blind review of the map, by a reader with none of the context: input, not orders.
7. Result page. A buyer conversation decides among the top few.

Onchain buyers

The account count is often public; read it instead of estimating it. Protocols: DefiLlama's Solana fee overview, protocols only, grouped by parent (worked count). Users: transacting wallets, deduplicated (traction checks), in humans or teams (crypto ICP). DAOs buy through proposals and grants, so their C is in past ones. Tokens convert at a dated price. Time or work saved instead of money: accelerates-x.md.

Output

• .claude/context/market.md, committed.
• A one-paragraph summary for the founder: the top wedge and why, the beachhead with its calculation, the proposed slide number, and the open decisions.
• On request, the wedge result page as one self-contained HTML file.