Use of funds

Where your pledge actually goes

Every dollar accounted for — or as close as a one-person studio can promise.

Most Kickstarter campaigns ask you to fund an idea. This one asks you to fund the next chapter of a game that already runs. Below is what we already built (on our own time), what your pledge actually pays for, and the priorities we promise to fund first.

What’s already real

Before you pledge a single dollar, here’s what already exists. You can verify each of these by opening arcomage.org and clicking around — the browser version is live, the AI opponent is live, the spellbook actually flips. No screenshots, no fake mockups.

The browser game. Open spellbook, flip pages, cast spells against a GPT opponent at arcomage.org/play.
The card-data pipeline. Spell content is data, not pixels. Adding a new spell is one JSON file + one illustration. Balancing is a one-character edit.
The Korge desktop engine. Eleven-phase turn system, two-player ply/round controller, spell-script API (Rhino-hosted), 60s/ply timer with spell-extendable cap.
Three hundred and eighty particle effects ported from the desktop client to the browser through Proton with 93 % automated coverage.
The AI architecture. AlphaZero-style policy+value network scaffolded, self-play loop running locally, Python service ready to scale.
Admin tooling. Player management, KS-bonus toggles, spell-creator wizard, audit logs — all already wired.
Two private GitHub repos backed up + secret-scrubbed. Nothing is on a single machine’s hard drive.

This isn’t a slide deck. Pledges are funding the delta between what works and what ships at scale.

Where the money mainly goes (priority order)

Funded in this order. If we under-collect, the lower-priority items get pushed to stretch goals, but the top four are the campaign’s commitment.

Guild & player Discord servers

Each Founder Guild (the $300+ tier) gets a hosted Discord channel + roster page on the site. Plus the general supporter Discord, the beta-update Discord, the dev-chat Discord, the public Discord, and the bots that connect them. Per-server hosting, bot infra, moderation labor.
Single-player campaign mode

A real story arc with scripted wizard duels, branching choices, and great in-campaign rewards. We want backers to have a thing to play through, not just a sandbox to duel in.
More spells

The launch deck is tens of spells; the multi-year vision is thousands. Illustration costs scale linearly with the spell count, and roughly nine out of ten generated spell ideas get rejected at review — so the real cost is closer to 10× the visible card count. We’re actively building templates and validators to drop that ratio, but it’s a real bill.
Multiplayer

Real-time WebSocket multiplayer at scale, anti-cheat (the server is the rules engine), matchmaking, ELO, replay storage. Multiplayer needs dedicated engineering hours we don’t have without funding.

The receipt — what each dollar actually buys

These are the real recurring + one-time costs we’d be funding. The numbers come from current measured spend — you can ask us to show invoices.

Neural-network training compute

The AlphaZero-style balance engine plays thousands of self-play games per night to surface which spells dominate the meta and which never get picked. That’s GPU hours, not free. Without funding, training caps at “whatever fits on the dev box” and the balance system becomes a marketing promise instead of a running system.

GPT multiplayer at scale

The AI opponent plays through OpenAI’s API. Per match burns API tokens; at hundreds of thousands of players playing millions of matches per month, this becomes a real recurring bill. We mitigate with caching, prompt tuning, and the in-house NN taking over the long-tail of matches, but the first year is the GPT bridge.

Spell generation pipeline

Each card’s illustration is generated at $0.04–$0.17 depending on quality. We use medium (~$0.04) as the production standard. About 1 in 10 generated spell concepts is actually shippable — the rest fail balance, art, or theming review. That ratio is the real per-card cost, and we’re building templates + a validator-driven prompt refinement loop to bring it down.

Hosting + infrastructure

arcomage.org needs to handle launch traffic, per-guild Discord servers need separate hosting + bot processes, the NN training replay buffer needs persistent storage, and the WebSocket multiplayer relay scales with concurrent matches. None of this is free at scale.

Developer salaries (mobile team only)

The only Kickstarter funds going to salaries are for a small team to build the native mobile app. Mobile play is on the roadmap (the browser version is already responsive), but a first-class iOS / Android app needs dedicated engineering. See the founder note below for the rest of the salary story.

Legal counsel

Card games adjacent to established systems need careful IP + licensing review, KS-compliance review, data-privacy review, ToS / EULA, age-rating, and any future cross-border distribution review. Lawyers aren’t cheap and we’d rather pay for one good review than discover problems after launch.

Founder note — no personal salary

Trent Tompkins (founder) is not personally taking a salary from Kickstarter funds. He has other income sources and is funding his own time. The only salaries paid from KS funds are for the mobile-app team described above.

This is a deliberate choice. ArcoMage is a project I’d build either way; what funding does is unlock the parts I can’t build alone — hosting at scale, training compute, engineering hours, lawyers. Backer dollars buy capacity, not lifestyle.

If you ever want to verify how the money is being spent: we’ll publish a quarterly transparency post on the supporter Discord with the running invoice totals + the priority milestones.

Ready to back?

View Reward Tiers Try the Browser Game First

Questions about the budget? Email us at the Discord or contact through arcomage.org — we answer.