A paid waitlist for the language buddy — get the first cash + 10 design partners, with the downside capped on purpose.
4 minTwo products. A Founders Pass — one subsidized year for 10 power users, to get cash + design partners — and a recurring monthly that's the real business and funds usage at steady state.
Trade-off: a subsidized first year, capped at 10 people
Trade-off: reasonable price now, tune margin later
Ship-ready — this is the exact block to hand off. locked
10 spots only
Fine print: unlimited use is subject to fair use — we may step in on genuine abuse. Founder pricing applies after the first year.
Keep "Unlimited beta use" as the visible bullet (clean, punchy); the fair-use caveat lives only in the fine print. The spots counter should reflect real signups.
This is the fact that drove every decision: an active user is expensive. ~$10 of inference burned in a few days of light testing → a realistic floor of $15–30/mo per active user.
That kills the obvious move. Healthy software wants cost to be ~25% of price — at a $15–20 floor that implies a $50–70/mo sticker. That's personal-trainer pricing, not Duolingo pricing. So you can't have all three at once:
Tiering and caps are how you escape the triangle later. For now, the free tier and the 10-person cap do the bounding.
Be honest about it: $99 for a year of unlimited use is below cost. A power user runs ~$180–360/yr, so you're subsidizing each founder by roughly $80–260. That's fine — it's a deliberate loss-leader, bounded three ways:
The real business model is the recurring $20/mo below — that's what funds usage at steady state. The Founders Pass just buys you 10 design partners and your first cash.
$99 upfront buys one year, unlimited, for 10 people. Here's the actual exposure for the year:
| Line | Per founder | All 10 |
|---|---|---|
| Upfront (1 year) | $99 | $990 |
| Est. cost to serve (unlimited, 1 yr) | ~$180–360 | ~$1.8k–3.6k |
| Net for year one | ~‑$80 to ‑$260 | ~‑$0.8k to ‑$2.6k |
| Bounded by | 1 year | 10 spots, then standard pricing |
Acquisition funnel — must be a rounding error
Reasonable now, re-price with real data
$20/mo is a starting number, not a verdict. At zero users you can't price precisely — lock it, watch real usage and cost, adjust at launch.
| Option | Why not |
|---|---|
| $1–5 refundable deposit | Too weak — filters for almost no real intent. |
| Lifetime unlimited | Sells your best users too cheaply forever while inference costs scale. The classic trap. |
| Locked $/mo for life | Permanent below-cost liability. Time-box the founder deal to one year instead. |
| Cheap subscription only ($7–10/mo) | Below the cost floor — loses money on every active user. |
| Tiers / credits / metered billing | Premature at zero users. Real problem to solve after there's usage data. |
| 100 founder spots | Can't honestly be on-call 24/7 for 100 people — the perk becomes a lie. |
Deliberately deferred — don't build these now, just know they're coming:
The waitlist's only job right now: prove someone will pay at all, and put 10 real users on your personal line. Everything else is a problem you want to have.