This document is private. Enter the password to continue.
Nine applications on a shared platform, wired into sixteen third-party services, for $339.52 a month. This is the cost side only.
Two kinds of number, never mixed. Documented figures are what we actually pay today. Modeled figures apply vendor list pricing, verified July 15, 2026, to the assumptions set out before the growth table. Anything that cannot be sourced to one of those is listed as an open item rather than filled in with a guess.
Every recurring technology cost, as of July 15, 2026.
| Service | What it does | Plan | Monthly |
|---|---|---|---|
| Claude | Primary development tool across all nine apps | Max | $200.00 |
| Supabase | Database, authentication, realtime, storage, server functions | Pro + add-ons | $40.52 |
| Vercel | Hosting and deployment for every web surface | Pro, 1 seat | $40.00 |
| Resend | Transactional email delivery | Pro | $20.00 |
| Expo EAS | Mobile app builds and over-the-air updates | Starter | $19.00 |
| Linear | Issue tracking and project management | Basic, 1 seat | $12.00 |
| GitHub | Source control and continuous integration | Team, 2 seats | $8.00 |
| Anthropic API | Analyst chat and automation keys | Prepaid credit | Usage |
| Sentry | Crash and error monitoring | Nothing collected | $0.00 |
| PostHog | Product and web analytics | Within free allowance | $0.00 |
| OpenAI | Content moderation | Moderation endpoint is free | $0.00 |
| Total | $339.52 |
| Item | Detail | Annual | Per month |
|---|---|---|---|
| Squarespace domains | Eight domains at $20.00 each | $160.00 | $13.33 |
| Apple Developer Program | Auto-renewing, ordered April 29, 2026 | $99.00 | $8.25 |
| Total | $259.00 | $21.58 |
| Item | Detail | Cost |
|---|---|---|
| Google Play developer registration | Paid April 7, 2026. Google charges this once and it does not renew, unlike Apple's annual membership. | $25.00 |
All in, the company spends $361.10 a month, or $4,333 a year, to run nine applications.
The most useful fact about the current bill is how little of it responds to usage.
Supabase is the only line with a variable component, and this month that component was fifty-two cents. The rest is subscription pricing that would be identical with a thousand players or none.
| Supabase, reconciled | Monthly |
|---|---|
| Pro plan base | $25.00 |
| Compute add-on, Small instance | $15.00 |
| Custom API domain | $10.00 |
| Less compute credit included with Pro | -$10.00 |
| Metered usage | $0.52 |
| Invoiced June 25, 2026 | $40.52 |
April's summary put the run-rate at $73, or $53 excluding Speed Insights. It is now $339.52. That gap is decisions and new tools, not drift.
Claude at $200 is the largest line and had no April equivalent. It replaced Cursor as the primary development tool. One subscription now does the work that the previous toolchain did, across all nine apps.
Two deliberate Supabase choices add $15 a month. The database was moved from the included Micro instance to Small in April, for development headroom. Net of the $10 compute credit Pro includes, that decision costs $5 a month, and it can be stepped back down at any time. Separately, a custom API domain was added at $10.
Expo EAS, Resend, and Linear became paid lines as the toolchain settled, adding $51 combined. Linear was free in April and is now a paid seat.
This one is worth explaining, because it does not look like infrastructure. Every room photo and tournament banner in the product is served from api.tblmaster.app, and those URLs end up embedded in pages, emails, and caches.
Pointing them at our own domain instead of a vendor-specific URL means the backend can be moved or re-pointed without breaking published content. That is the real return: it buys vendor independence for $10 a month. Keeping our project identifier out of client requests is a fair secondary benefit, though modest hardening rather than a control on its own.
Everything past this point is calculated. These are the inputs, and the right thing to challenge.
Five stages, anchored to the room directory: 485 poker rooms confirmed across 41 states.
| Stage | Rooms | Active players | Fixed | Usage-based | Total monthly | Per room |
|---|---|---|---|---|---|---|
| Today | 0 | 0 | $339 | $1 | $340 | - |
| Pilot | 1-3 | ~1,000 | $339 | $1 | $340 | $113.33 |
| Florida | ~28 | ~10,000 | $365 | $86 | $451 | $16.11 |
| Regional | ~150 | ~50,000 | $425 | $698 | $1,123 | $7.49 |
| National | ~485 | ~170,000 | $754 | $2,923 | $3,677 | $7.58 |
Today's row is documented · every other row is modeled
The pilot is free. One to three rooms and a thousand players fit inside allowances we already pay for, so the first live deployment costs nothing beyond today's bill. Meaningful cost starts around the Florida tier and stays modest well past it.
These figures price the platform built and operated properly, because that is how it will be built and operated. It will still vary, and the reasons are worth naming.
Players per room is the dominant lever. Every usage-based line scales off the assumption of 350 monthly actives per room. The variable component is close to linear in player activity: if rooms run 50% busier than modeled, the national figure moves from roughly $3,677 to about $5,100. That is a revenue event as much as a cost one.
Engagement depth moves analytics and transfer together, since a more engaged player generates more of both. Feature mix matters too: realtime-heavy additions raise database and transfer costs faster than page-based ones.
Database compute is a ladder, not a slope. Supabase steps at $15, $60, $110, and $210. Crossing a step is a jump rather than a drift, and where we cross depends on query patterns as much as on volume.
Vendors change their pricing. Google withdrew its $200 monthly mapping credit in March 2025. That kind of change is outside our control and is the reason to re-verify this model annually.
None of it sits unwatched. Analytics and error monitoring are instrumented today, and both Supabase and Vercel expose usage dashboards and spend caps. Cost drift surfaces within days and gets corrected, rather than compounding quietly across a quarter.
Sixteen third-party services, but only four ever become material.
Everything else stays small. Builds, source control, issue tracking, and development tooling total $239 a month and do not change with scale, because the model assumes a solo operator throughout.
Real costs, not yet quantified. Listed rather than estimated.
Real costs, deliberately excluded, noted so they are not mistaken for oversights.
The platform costs $361 a month all in, and almost all of that is fixed. Nine applications, sixteen integrated services, and a national data pipeline run for less than a phone plan, and the first pilot rooms are absorbed without the bill moving at all.
Growth does not break that. It costs roughly two cents per active player per month, and the cost per room falls from $16 in Florida to about $7.50, where it flattens. Even at all 485 confirmed poker rooms in the country and 170,000 active players, the modeled infrastructure bill is under $3,700 a month.
The real uncertainty is not vendor pricing. It is that no assumption here has met a live player yet. The pilot is what converts this from a model into a measurement, and it costs nothing to run.