Market: how to read the rings.

TAM is total addressable market: every dollar that could be spent on this category by everyone, everywhere, if there were no friction. It tells you whether the problem is large enough to matter. SAM is the slice you can serve with your current product, geography and channel — same problem, narrower lens. SOM is what you can realistically capture in three years given your team, money and execution speed.

The mistake is using TAM as a proof of seriousness ("$100B market!") and SOM as if it were guaranteed. The correct reading is inverse: TAM proves the category isn't a hobby; SOM proves you've done the maths. A founder who can't compute SOM bottom-up is asking you to trust their vibes.

Read competitors in three layers. Direct: same job, same channel — they reveal your real price ceiling and feature floor. Adjacent: same job, different shape — they teach you which assumptions to challenge. Substitutes: not products at all — the hack people use today (a side tab with Google Translate, a teammate translating inline). Substitutes are the truest measure of pain.

The worst category is one with many adjacents and zero direct — it usually means the problem is too narrow to fund a dedicated tool. The best category is one with several well-funded direct competitors all attacking from one side and ignoring another. The ignored side is your wedge.

Pick two axes that are independent — not correlated with each other and not a restatement of price-vs-quality. Plot every direct and adjacent competitor. The valuable output is not where you sit; it is the quadrant that stays empty. An empty quadrant is either a wedge (no one has solved it yet) or a graveyard (no one wants it). The interview script in Lesson 02 tells you which.

Two warnings. First, do not invent axes to put yourself top-right; investors see this in seconds. Second, an empty quadrant is a hypothesis, not a moat — Lesson 06 explains how to defend it before Big Tech notices.

For every trend you write down, finish this sentence: "because of this, we will do X and not Y." If you cannot finish it, the trend is decoration. Three buckets matter: technology curves (what becomes possible), consumer behaviour (what becomes habitual), and regulation (what becomes mandatory). Each demands different proof — benchmarks for tech, surveys for behaviour, statutes for regulation.

Founders fail at this in opposite directions. They either invoke trends abstractly ("AI is changing everything") or worship one trend so hard they ignore the others ("on-device latency dropped, therefore we're a winner"). The market sits at the intersection of all three. So does your decision.

Wedges expire. The minute your gap is visible — through a Product Hunt launch, a TechCrunch piece, an investor announcement — the clock starts. Moats grow inside that window: distribution lead, integration depth, data flywheel, brand. A founder who confuses "no one has done this" with "no one can do this" is borrowing a moat they have not earned.

The honest read on most software moats is brutal — model quality is commodity, APIs are interchangeable, and design is copyable. What is not copyable in twelve months: cumulative OS-integration work, a base of paying habits, and a creator community that names you as the default.