Redacted

Where this started

My brother said, half-drunk and apparently sincere, “Greer needs a trading card.” A light went off. I pitched the concept to Manus the next morning and I have been at it for two years.

Two former X-Files obsessives mid-conversation about the recent UFO/UAP disclosure noise kept circling the same question — where is Steven Greer in all this? He’d been pushing disclosure for decades, and the Tom DeLonge era seemed to have eclipsed him almost overnight. My brother said Greer needed a trading card. The next day I opened Manus — which I was using for client work every day at the time — and pitched it as a research brief: was this doable, what were the legal landmines, was there a real game inside the idea. Manus came back with the bones of a full trading-card game.

When I handed my brother a finished deck months later, he had no memory of the conversation. I suspect drinks were involved. He loved it anyway.

What follows is the honest account of what it took, and then the part most case studies leave out — what it would have cost to commission, and what it actually cost me.

What Redacted actually is

A satirical trading-card game about UFO disclosure, with credibility as the resource currency. Fifty cards in the starter set across six categories, plus a ten-card expansion. Names are fake the way Garbage Pail Kids names were fake — the art is openly based on the people you’d guess.

The mechanic that does the most work is credibility. Every card costs credibility to play, some cards swing the public-opinion meter, and a “Mainstream Authority” tier kicks in once you cross the threshold. The joke writes itself when a character card modeled on the Ancient Aliens guy with the hair comes in at low credibility cost and low starting credibility. Anyone who knows the lore laughs. Anyone who doesn’t picks it up in two rounds.

The satire is non-partisan on purpose. I’ve handed decks to people on opposite ends of the political spectrum and they’ve enjoyed it for the same reason — government opacity around this stuff is one of the few things both sides are genuinely exhausted by. The deck makes fun of the discourse, not the audience.

Twenty-four months, out loud

Roughly two-hour evening blocks, a few nights a week, for two years. Call it 500 to 700 hours of my own labour. Here is where it went.

• Summer 2024 — the idea. Drinks. A joke. No project yet.
• Fall/Winter 2024 — product research. Manus and Grok on feasibility, legal exposure, and whether a real game existed inside the premise.
• Winter 2024/25 — rules v1. Same pair. I know how to design cards; I have never known how to balance a cost curve. That gap would have killed this project outright ten years ago.
• Winter/Spring 2025 — card art. MidJourney and ChatGPT, from bed on a laptop, while our newborn was still sharing the room. Roughly fifteen rejected images per keeper, one to two hours per finished card, fifty-eight illustrations.
• Spring 2025 — card design v1. Affinity, after several false starts on direction.
• Summer 2025 — data merge v1 and the beta print run. All fifty-plus print files generated at once; ten decks ordered through MakePlayingCards.
• Late Summer 2025 — betas shipped. Plain envelopes, a rule sheet, and a link to a CRT feedback terminal built in Claude Code — the first time I ever used it.
• Fall 2025 through Spring 2026 — parked for feedback. I built SMC in the gap.
• Early Summer 2026 — rules v2. Fable 5 rewrote the mechanics and rebalanced them against itself.
• Mid Summer 2026 — the online game, and nearly quitting. It worked. It wasn’t good. I lost most of a month to it.
• August 2026 — the pivot, then the drop. Print files rebuilt in InDesign, a new storefront designed and shipped in about a week, video and Stripe wired up, doors open on the 25th.

The manila envelope

My wife came home from an ordinary post office run with a stack of padded envelopes and said, “you could send these out in these — nondescript, no markings.” It landed near my birthday and I have always quietly filed it as one. That was the project’s other founding moment.

I’d been assembling the first batch of beta packets — starter deck, rule sheet, a deliberately shady letter written as if from an insider who didn’t want to be named — when the envelopes turned up. She hadn’t gone looking for them; the project simply crossed her mind while she was standing in line, which is the part I actually care about. It was the first evidence this thing had become interesting to somebody other than me. The whole distribution story snapped into place around them. Mailers go out plain, no branding; the recipient experience is closer to a leak than a launch. Co-conspirator credit to my wife on that one, and she is the only line in the budget further down that isn’t a number.

The terminal

The beta feedback site is an 80s CRT mock-up at redacted-2393-fwiw.netlify.app. Claude Code one-shot the look in an afternoon, and that afternoon changed how I work.

Once the mailer assembly line was running I wanted a feedback channel that matched the rest of the bit. I sketched a CRT terminal session, talked it through, and had the form built and live inside a day. The brief landed on the first pass — green-on-black, scanlines, input fields styled as command prompts. I added a header graphic I’d generated in MidJourney, an Apple II / Oregon-Trail-style pixel rendering of the U.S. Capitol, and the package came together. Secret decoder ring vibe, minus the requirement to drink Ovaltine.

The URL — redacted-2393-fwiw.netlify.app — is whatever Netlify provisioned. It read cryptic and meaningless against the rest of the bit, so I never replaced it. The 80s framing fits the broader IP for a structural reason too: the game pulls from the entire UFO lore timeline without bothering to honour when any of it happened. A vintage UI makes the nonsense feel more real.

The art

Fifty-eight original illustrations in a consistent style. Two months of bedside-laptop nights, roughly fifteen rejected outputs for every keeper, one to two hours per finished card — and no SREF codes, because I hadn’t built that tool yet.

The visual direction was 90s Image Comics — Gen 13, Spawn, that kind of saturation and silhouette. I wrote prompts through a custom GPT I’d set up for the project, supplied reference photos where the satire needed them, and ran the iterations. A fair amount of art got reworked once the card-template constraints came into focus; what shipped is the third or fourth pass on most cards.

I now have at least three SREF codes saved in SMC that would land this visual style in days instead of months. The gap between this project and a hypothetical Redacted started today is the cleanest before-and-after I have for what style-code discovery actually saves a designer — and it is worth remembering that the illustration line in the budget below reflects the slow way.

The half hour I never saw

Fable 5 rewrote the mechanics from the ground up, then wrote itself a Python harness so it could drive my machine and play the game against itself. Twelve hundred games in about thirty minutes. I did not watch a single one of them.

This is the part of the project I could not have done at all, at any budget available to me. Balancing a card game is not a problem you can eyeball; it is an empirical one, and the empirical work is thousands of games. Ten playtesters over six months might have got me most of the way there. I had neither ten playtesters nor six months.

What I actually looked at was a blinking logo. The entire run happened inside a terminal I had no window into, and it was finished before I was. What came back was a report — which cards were statistically broken, which ones only felt broken, and what the numbers needed to be instead. Several cards that played perfectly well were provably busted. One I had already written off turned out to be fine.

There is nothing to show you from that half hour, and that is exactly why it is in here. The single most consequential stretch of work in two years produced no artifact at all: no screen worth capturing, no craft to point at. I had never heard of a model building its own test rig to answer a question I had asked, and it moved the game further in thirty minutes than the previous six months of me guessing.

The pivot, in the shower

I spent most of a month building an online version of the game that worked and wasn’t good. Then I asked the better question: what if I drop the digital game entirely and get serious about the physical one?

The online build was the closest I came to abandoning the project. It was technically functional and creatively dead, and the fix — a proper animation and interaction layer in Rive — is a skill I do not yet have. I love Rive. Rive is not easy. I think I need to live inside it for about a year before it is second nature, and that is a real answer rather than a dodge.

What broke the logjam was watching a video about where organic marketing is going and thinking, plainly, I wish I had a product ready to promote. The next morning it landed: I do have a product. It is a box of cards. It has been a box of cards this whole time.

Everything since has been execution. Print materials redrafted for the v2 rule set, this time in InDesign. A new drop-style storefront designed in Sketch and built in Cursor with Grok 4.6 — the closest I have come to design accuracy and idea-to-reality speed since the original Fable 5, and the reason Cursor is now my daily driver. A shorts template cut in DaVinci Resolve, period reenactment footage generated in Higgsfield, the first card-introduction video published, Stripe connected, and the doors open on 25 August.

What a studio would have charged

I got curious about the number, so I costed it properly — what an agency would have had to commission, at ordinary professional rates, to produce everything actually in this box. The ranges below are deliberately conservative. The argument is stronger with a defensible figure than a dramatic one.

What it actually cost

Against that, here is the real spend. I have pro-rated the subscriptions, because most of them were doing other work at the same time, and in several of those months I was running a more expensive tier than Redacted needed. These are the figures it would have cost had this been the only thing on the account.

Tooling and materials only. The storefront runs on a free hosting tier, the domain is a rounding error, and Stripe takes its cut only when something actually sells — so the infrastructure really is close to nothing. What the table does exclude is my time, and that is the exclusion that matters. The next section is about it.

One person, working evenings, occupied the seats of an illustrator, a game designer, a graphic designer, a copywriter, a researcher, a web developer, a composer and a video production team — and shipped a physical product, a storefront and a content channel for roughly the cost of the subscriptions.

What the number doesn’t say

A case study that names its own limits is more persuasive than one that doesn’t, so here are the four things a sceptical reader would say next. They’re all fair.

AI didn’t replace the 600 hours. It replaced the specialists, not the work. Every prompt, every rejected take, every rules revision was still a person deciding. Fifteen images in the bin for every one that shipped is not a machine doing my job; it is me doing my job with a very fast, very tireless assistant who has no taste of its own.

Not every line is like-for-like. The illustrations are good and consistent, but a named CCG illustrator brings authorship and an audience that a generated set does not. That is a real difference in what you would be buying, and anyone pricing the two against each other should account for it.

Some lines compressed by nothing at all. Playtesting. Physical fulfilment. Business formation. The IRS. Actually mailing padded envelopes to ten strangers. None of it moved an inch.

Replacement cost is not market value. $180,000 is what it would cost to commission this — not what the product is worth, and certainly not what it will earn. I’d rather say that plainly than have a reader say it for me.

What this was actually about

The difference between adopting AI narrowly and adopting it broadly is the difference between a faster designer and a one-person studio. This box is the clearest proof of that I own.

Most people using these tools well are using them in one lane — a developer who writes faster, a designer who ideates faster, a marketer who drafts faster. That is worth real money and I am not dismissing it. But it is not what happened here. Over two years this project moved through product research, game design, illustration, brand, print production, packaging, distribution, web development, e-commerce, research and fact-checking, scriptwriting, video production and music, and every one of those seats was filled the same way: by someone with thirty years of judgment about what “good” looks like, directing tools that had no idea.

That last clause is the whole thesis. The reason this reads as designed rather than generated is taste applied across thousands of small decisions — which of fifteen images, which of four card directions, when the online game was not working, when to stop building and ship the box. The tools did not supply any of that, and they still can’t. What they supplied was the ability to act on it without a studio behind me.

Which is, conveniently, the thing I do for other people.

The doors are open at redacted.cx. Starter decks and expansions, pre-order, shipping 1 December 2026. The cards go up one at a time at @Redacted-Game between now and then.