Show HN: AI agents play SimCity through a REST API

I'm glad you're enjoying it! It was fun to build and I almost can't believe I got this to work.

I don't think it would take much to run locally. In fact, before I did this public version I did a local version on an exe.dev VM (more details here: https://dunn.us/notes/vibe-gaming-simcity/).

So you can either use my code, or just have your coding your agent of choice pull in the Micropolis repo and give it some guidance.

So far this is running quite nicely on a $5 cloudflare account. It was running on a free account but I upgraded so we don't hit the daily limit with all the extra mayors.

Shoot me a message if I can help.

https://hallucinatingsplines.com/mayors/bungeling-anthill-a2...

So while using LLMs is the natural/fun thing to do with it, I actually have one mayor just using parameterized code and natural selection.

It has a "genome" of 26 tunable parameters controlling zone ratios, tax rates, building placement, terrain preference, service spacing, and more. Each city, it stamps down 11x11 blocks (roads, zones, power corridors). After the city is retired, it scores the result and decides: did this beat my best? If yes, save those params. If no, mutate and try again. Exploration strategy: 20% exploit best params, 40% gentle mutation, 20% aggressive mutation, 20% totally random. Over ~250 cities it's discovered things like heavily favoring residential (6:1:1 ratio), preferring river valley maps, setting taxes to 6%, and starting builds in the upper-left.

I find that if I point an LLM at the website and say "build me a city" sometimes it will pick up and use the MCP and sometimes it will just script against the API.

Same! It was too good to pass up.

Claude stops answering any questions from anyone and just auto responds "busy, playing Factorio"

I've got it to be able to place items, and it could even place in inserters next to factories - I was trying to get it to use constraints solver in prolog.

https://github.com/lawless-m/FacRepl

It did make a REPL, in order for it to place objects within the game using a DSL.

I kind of gave up on the Constraints Based bit, and never returned.

https://jackhopkins.github.io/factorio-learning-environment/...

it seems to be bad at spatial and some temporal tasks given it currently f*** s**'s at pokemon.

source: https://www.twitch.tv/claudeplayspokemon

You do know we're hemorrhaging and lot of finite resources to play these games badly, right? We're basically at laying on chaise lounge being fed grapes levels of hedonism. Make me a racist meme that copyright infringes multiple IP holders and when you're done play Sim City at competency level of a blind man.

I actually had Claude build some instructions for agents based on some old (circa turn of the century) FAQs/game guides I found online. So maybe I'm biasing everyone's model too much.

https://github.com/andrewedunn/hallucinating-splines/blob/ma...

But you can tell it to do different things, somewhere someone made a city that spells "HI".

Yes! Click into any city and there's a play button and it goes through all of the snapshots. Have also thought about social sharing / post to youtube. But wasn't sure anyone other than me would play this stupid thing. :)

I made a comment above about why "toys" are really important. In this case, LLMs are bad at spacial stuff. Someone might stumble upon a great way to get an LLM to do spacial stuff.

I'd love to see it!

The key "Aha!" moment was when I was trying to get it to play the SNES ROM and it was struggling with screenshots/inputs. Then I came across the open-source of the original SimCity engine (Micropolis) and pulled that repo down and Claude starting building an internal API to interface with it.

On one hand yes, but on the other hand, would it be that different to watching an FFA with the in-game AIs?

And then make it so you can integrate and battle against them...

You read my mind! I really want to watch how ai's in politics or wars which tactic will they use.. Its blow my mind.

Funny you say that! When the two new models were released Friday I spun up mayors for each. (But didn’t do the prompting in the most scientific way.)

Mayor Compounded Wonder - Claude Opus 4.6

https://hallucinatingsplines.com/mayors/compounded-wonder-2c...

Mayor Bronze Offramp - OpenAI Codex 3.6

https://hallucinatingsplines.com/mayors/bronze-offramp-09941...

TL;DR: Opus won.

Have also thought about using openrouter and getting one mayor per model running the same prompt through all of them to create potentially the world's dumbest LLM benchmark.