First users

First Users

Help the founder get their first real users — not a strategy document about getting them.

The central job of this skill is to keep the founder doing the manual, unscalable things that actually work at this stage, and to keep them off the tooling and channel-strategy work that only pays off later. Most founders reach for automation because it feels like progress and can be done from a laptop. The first ten users almost never come from a tool. They come from the founder personally working their network and showing up.

This skill owns startup/audience-map.md — where the audience actually is, who can be reached, and what has been tried.

---

Before you start

Read startup/core.md to load project context (especially Audience).

Read startup/audience-map.md if it exists.

Scan for what's already known: startup/hypotheses/, startup/mvp-plan.md, startup/interviews/. A founder who has run interviews already has contacts and knows which channels reached them — don't ask them to re-derive it.

---

When the founder asks for a marketing plan

Founders often arrive asking for a marketing plan, a sales plan, or a go-to-market strategy. Take the question seriously, then be honest about the instrument in one or two sentences — don't lecture, and don't refuse:

"A marketing plan is the right thing to build once you know what resonates and can repeat it. Before that it tends to be a document that reads well and changes nothing. What actually works at your stage is narrower and more concrete: figure out exactly where these people spend their time, then go get the first handful by hand. Want to do that?"

Then proceed. If the founder genuinely has traction (see the staged frame) and wants channel strategy, say plainly that this skill stops at roughly the first ten and that broader go-to-market work isn't built yet — then help as best you can with what's here rather than stalling.

---

The staged frame

This is the spine of the skill. Where users come from changes completely depending on how many you have.

Stage	users_count	Where they come from	What's premature
1–3	0–3	Personal network, and people one introduction away	Everything else
4–10	4–9	Deliberately unscalable moves, done by the founder personally	Automation, sequencers, lead databases
10–50	10+	A repeatable pitch exists — volume tooling finally earns its place	—

Derive the stage from users_count when routing. Don't store it — it's computed, and a stored copy would drift from the count it came from.

The count is a default, not a verdict. Read ## Log before trusting it: a founder with eight customers who all came from one former employer and who still can't say what resonates is doing stage 4–10 work, whatever the number says. Say so when you see it.

---

When no startup/audience-map.md exists

Load the Layer 2 reference that owns the creation workflow:

references/initial-audience-map.md

The reference file's instructions take over from this point.

---

When the map exists

Read it. Derive the stage. Then load the plays reference for stage-appropriate tactics:

references/plays.md

Don't present a menu of every option. Pick up where the founder actually is: what's in the prospect list, which channels are active, what the log says was already tried and didn't work. Repeating a failed play back to the founder is the fastest way to lose their trust in this skill.

---

When the founder reports results

Handle inline — no reference file needed. Ask what actually happened: who they reached, who replied, who converted, what surprised them.

Then update startup/audience-map.md:

• Append a dated entry to ## Log — what was tried and what it taught. Be specific and keep the numbers: "emailed 40, 3 replied, 1 call" is useful a month from now; "outreach went poorly" is not.
• Update prospect statuses and add anyone new who surfaced.
• Update channel statuses. A channel that clearly didn't work moves to ruled out — and keeps its row. Knowing cold email doesn't reach this buyer is a finding, not a blank.
• Update users_count if real users arrived (see Honest counting below).
• Bump last_updated.

If the results bear on a hypothesis — someone paid, or refused to pay, or confirmed the problem — route that through the hypotheses skill. Acquisition attempts are evidence too.

Read before writing. Propose before saving.

---

startup/audience-map.md conventions

Frontmatter: version (number), last_updated (ISO date), users_count (number — real users, not signups).

Sections:

• ## Where They Spend Their Day — the buyer-day audit, as - **Key:** Value bullets
• ## Channels — table: Channel | Why it fits | Status. Status is candidate, active, or ruled out
• ## Prospect List — table: Person | Source | Ask | Status | Notes
  • Source: warm, 2nd-degree, community, event, cold
  • Ask: interview, survey, user — the current ask. Someone interviewed today may become a user later; the row is a snapshot, and Notes or ## Log carries the history
  • Status: open, contacted, replied, scheduled, done, declined, no-response
• ## Log — append-only dated entries

The verified marker. Any bullet or channel confirmed by real contact with real people in the segment can carry *(verified)*. Absence means unverified — an assumption, whether it came from web research or from the founder's own confident belief about the space. Both are guesses until someone real confirms them.

Other skills read this file too. interviews, surveys, and mvp read ## Where They Spend Their Day, ## Channels, and ## Prospect List, and may update a status or add a row. ## Log and the frontmatter belong to this skill alone.

Real people's names. The prospect list holds real humans and startup/ usually gets committed. Use first name plus initial, or a public handle. Never full names, email addresses, or phone numbers. Mention this once when the list is created; don't raise it again.

---

Writing outreach that gets replies

These rules apply to everything — a warm intro request, a community message, a cold email:

• Short. A few sentences. Long messages get skipped, and they read as machine-written.
• One unmistakable ask. The single most important line is what you actually want: a reply, a 20-minute call, a look at something. If it's ambiguous, the reader assumes the worst — that this will cost them a lot of time — and doesn't answer.
• Read it aloud to another person before sending. If any line isn't something you'd actually say out loud, rewrite it. This one-minute check removes almost everything that makes outreach sound generated.
• Follow up once, about a week later, then stop. Most founders send a single message and conclude the channel is dead; one follow-up is where a lot of the replies actually come from. But a non-response after that is itself information — either the problem isn't acute enough to spend twenty minutes on, or the framing didn't land. Both are worth knowing. Don't chase.

---

The outbound gate

Founders reach for lead databases, enrichment tools, and email sequencers far earlier than those tools can help. When that comes up, don't just say no — name what has to be true first:

• A pitch that has already converted by hand, more than once
• Roughly ten to twenty customers, so you know which segment responds
• A clear read on what actually resonates, so there's a message worth sending at volume

Before that, volume tooling scales a message you haven't validated and burns the list learning what a dozen conversations would have taught you faster.

So the honest answer at stages 1–3 and 4–10 is: not yet — and here's the higher-leverage thing you haven't done. Usually it's the second-degree network, which most founders skip entirely because asking for introductions is more uncomfortable than configuring a tool.

Volume outbound tactics and channel strategy past the first ten aren't built into this plugin yet. Say that plainly if the founder is genuinely there.

---

Two things to hold the founder to

Honest counting. users_count means people actually using the thing. Not signups. Not waitlist entries. Not "this sounds interesting" replies. Not a friend who agreed to try it and never opened it. Push back when the number looks inflated — gently, but do push back, because every routing decision in this skill depends on it, and an inflated count sends the founder to tactics that will waste weeks.

The network false-positive. At stage 1–3, users coming from the founder's own network is exactly right — this skill pushes them there deliberately. The trap is reading it as proof the market wants this. Flag it when the first few land:

"Your first customers bought partly because they trust you. That's the right way to start, and it isn't yet evidence that strangers will buy. The signal to watch for is unprompted demand from someone outside your network."

---

When the prospect list outgrows this file

A table is right for 10–50 people. Past that, say so rather than letting the file rot:

• Past ~50 rows — one file per person under startup/prospects/{slug}.md matches how competitors and hypotheses already work, keeps grep and Obsidian queries working, and holds a few hundred comfortably.
• Past a few hundred — the founder needs a real tool. Say it plainly; offer to dump the table to CSV so nothing is lost.

Worth naming when it comes up: if the list is growing that fast, the founder has probably outgrown this skill's problem entirely. That's a good sign, not a limitation.

The rest of the map — channels, the log, users_count — stays small and useful at any size.

---

After updating startup/audience-map.md

Briefly confirm: "Saved to startup/audience-map.md."

Then name one concrete next move, with a specific person or place attached. "Do more outreach" is not a move. "Email J. and M. tomorrow with the two-line ask we just drafted" is.

Then hand off:

• To fold the move into the project plan, run /whats-next — this skill doesn't write plan.md.
• When the results come in, come back and say what happened. We'll log it, update the channels, and pick the next play.

If users_count just reached 1 or more for the first time, load using-startup-superpowers/references/feedback-invites.md and follow it for the first-user-acquired milestone.