Spending my last 6% of Fable on the tails of the bell curve

My Fable use was down to 6% before it gets paywalled. I'd been throwing everything I had at it and been impressed the entire time. So for the last of it, I pointed it at Polis.

Polis is my political-negotiation game: you play the mayor of an ancient Greek city-state, and the twist is that the faction leaders are LLMs you negotiate with in live conversation — the terms you settle on get parsed into structured deals that a deterministic simulation engine actually enforces. I'd gotten it to a working state, but I had stepped back from it: the concept is interesting, but it's still missing the gameplay to make it cool. (Also graphically uninteresting.)

Nate B Jones made a point that stuck with me: most LLMs are good in the middle of the bell curve. Fable can try things out on the tails. So that's exactly what I asked for:

"Projects\Polis Please look at this spec and design (Not code) I want your ideas on it. What do you think would be interesting to add/change. High concept. As out there as you can. this is brainstorm no bad ideas"

It gave me pure gold. Some were half-ideas already in my head; some were things I never would have thought of. All of it got the creative juices flowing.

Then I prompted it about the visuals, gave it a couple of concepts of mine, and it blew my mind again. I'm not a graphic artist — it riffed on my ideas, then explained which tools were actually right for building it and why. I had it write the whole art direction up as a document for the project.

Then one more stretch:

"Any ideas overall for the project. Time to stretch the mind. What crazy things should I do with this concept?"

Again, great ideas. Most will never happen — but that's not what brainstorming is for. Where my head is at on this project now versus a week ago is the point.

The whole thing took 4 of my remaining 6%. Fable is a great idea model.

** Highlights below; full conversation attached.

4. The Oracle — shadow simulation as prophecy. (My favorite idea in this list.)

You have a deterministic engine that runs a full cycle suite in ~2 seconds, stub mode, zero dependencies. So: fork the current snapshot, run it forward 5–10 cycles headless, and hand the actual trajectory to an LLM instructed to speak it as deliberately ambiguous Delphic verse. "The wooden wall will save the city" — the prophecy is true, because it's a real forecast, but it's the player's job to decode it. This turns your election-forecast prerequisite ("a visible approval needle") into an in-fiction mechanic, and it's a feature no other game can copy without your architecture: prophecy is only honest if the engine is deterministic.

14. Dynasty mode.

A run is a reign; a campaign is a city across generations. On game over, time skips twenty years and you return as a new Mayor in the same city, scarred — projects decayed, factions remembering your predecessor ("the Grain-Hoarder," the historian's epithet from idea 5, now in the audience prompt). New-game-plus where the "plus" is reputation sediment.

From the Out there prompt

9. The Eternal City.

One persistent Polis running forever on your isolated box, cycle per hour, publishing its chronicle as a website/RSS feed. Visitors get one audience each — a stranger's deal ripples through a city thousands of others are watching. It's slow TV, an art piece, and a permanent live demo of your whole stack that costs you nothing but electricity.

1 comment

Spending my last 6% of Fable on the tails of the bell curve