May 30 2026

I ❤️ OILBARON.EXE

I’ve been on a weird retro gaming kick lately. I played through full runs of all of Doom and Doom II, rocked out to some Heroes of Might and Magic III, tried playing Tie Fighter but it just wasn’t the same without my old joystick. I also tried command and conquer red alert, the xcom that takes place under water whatever that was called and even loaded up some of the first civ games. Pure nostalgia. I guess I’m an old man now.

There was one shareware game that I had a vague memory of that took me a while to find. Oil Baron was a tile based game I played when I was maybe 14 years old?

OILBARON v1.2: Oil Baron Game For Windows Oil Baron is a board-type strategy game for Windows. Build companies on a grid, trade shares, cause mergers. Play with up to six players, inc. the computer. V1.2 adds right button functionality, new statistics windows, more.

And because it’s a win16 based game I couldn’t get it running without the help of a cool little project that adapts wine to allow running 16 bit windows apps on 64bit windows via emulation. I installed it via this installer https://mendelson.org/otvdm.html. Once installed I was able just to directly run the OILBARON.EXE with no issues. By far the simplest solution vs emularity or other emulators I chased. It

Stock Baron - a web based clone

Anyway this post is not specifically about getting the game to run, but about the clone I built in homage to Al Funk’s little masterpiece. I almost never run Windows and I wanted the ability to play this game from basically anywhere, so I set out to make a web based clone. I targeted TypeScript and HTML with the intention of building a version of the game that was fully portable and ideally runnable offline. I specifically didn’t want a server involved to keep distribution costs at zero as I have no intention of commercializing what is effectively a fan homage.

You can play the end result here: https://stock-baron.com

The Process

I created this clone in the span of an afternoon roughly and have subsequently refined it a few times based on my own desire for little features (configurable games, graphs) and improved playability (stronger bots, less clicks), but the bulk of the project was completed in 4-6 hours which is yet another mind numbing experience with Claude at my side.

Here’s a retroactive log of how I approached it and where I ended up:

Project Log

PROMPT

I’m writing a blog post about this project and want to articulate the work we did in those first two sessions (83 and 79 msgs). Can you provide a bulleted list of steps we took over those two sessions?

Note

I’ve bolded the steps in Claude’s response where we had the most fun.

The act of playing a game I had just made (cloned) against an AI was surreal. The AI as usual was padding my ego and complimenting my strategy with full on sycophancy, but at the same time was making genuinely insight observations about my strategy that it was then able to bake into the more deterministic bots that ship to the client.

Here’s what the CLI version of the game looks like. I have an alias to the typescript program called “sb” and every sb invocation reads the game state, presents the board and waits for your input. The game commands are the same commands that can be used in the web version of the game if you prefer staying on the keyboard as I do.

stock-baron ❯ sb new

> [email protected] cli
> tsx src/cli.ts game.json new


        A   B   C   D   E   F   G   H   I   J   K   L
  1    .   .   .   .   .   .   .   .   .   .   .   . 
  2    .   .   .   .   .   .   .   .   .   .   .   . 
  3    .   .   .   .   .   .   .   .   .   .   .   . 
  4    .   .   .   .   .   .   .   .   .   .   .   . 
  5    .   .   .   .   .   .   .   .   .   .   .   . 
  6    .   .   .   .   .   .   .   .   .   .   .   . 
  7    .   .   .   .   .   .   .   .   .   .   .   . 
  8    .   .   .   .   .   .   .   .   .   .   .   . 
  9    .   .   .   .   .   .   .   .   .   .   .   . 
 10    .   .   .   .   .   .   .   .   .   .   .   . 
 11    .   .   .   .   .   .   .   .   .   .   .   . 
 12    .   .   .   .   .   .   .   .   .   .   .   . 

Corps: (none active)

Player 1 (Human) $6000 cash:$6000 | (no shares)  ← current
Player 2 (Elon) [bot] $6000 cash:$6000 | (no shares)
Player 3 (Sacks) [bot] $6000 cash:$6000 | (no shares)
Player 4 (Chamath) [bot] $6000 cash:$6000 | (no shares)

Phase: place_or_discard | Round 1 | Turn 1 | Chits left: 116
Your chits: [H1] [H3] [G7] [E11] [E1] [K9] [G1]

Game saved to game.json

Next turn after placing my first chit and the bots auto-playing their turn (sb place G7 then sb end):

> [email protected] cli
> tsx src/cli.ts game.json end

(ran 8 bot actions)

        A   B   C   D   E   F   G   H   I   J   K   L
  1    .   .   .   .   .   .   .   .   .   .   .   . 
  2    .   .   .   .   .   .   .   .   .   .   .   . 
  3    .   .   .   .   .   .   .   .   .   .   .   . 
  4    *   .   .   .   .   .   .   .   .   .   .   . 
  5    .   .   .   .   .   .   .   .   .   .   .   . 
  6    .   .   .   .   .   .   .   .   .   .   .   . 
  7    .   .   .   *   .  OA  OA   .   .   .   .   . 
  8    .   .   .   .   .   .   .   .   .   .   .   . 
  9    .   .   .   .   .   .   .   .   .   .   .   . 
 10    .   .   .   .   .   .   .   .   .   .   .   . 
 11    .   .   .   .   .   .   .   .   .   .   .   . 
 12    .   .   .   .   .   .   .   .   .   .   .   . 

Corps: OpenAI($7.50, 2 tiles)

Player 1 (Human) $6000 cash:$6000 | (no shares)  ← current
Player 2 (Elon) [bot] $6000 cash:$6000 | (no shares)
Player 3 (Sacks) [bot] $6000 cash:$6000 | (no shares)
Player 4 (Chamath) [bot] $6140 cash:$65 | OA:810

Phase: place_or_discard | Round 2 | Turn 5 | Chits left: 112
Your chits: [H1] [H3] [E11] [E1] [K9] [G1] [F1]

A new corp (OpenAI has been formed!)

You get the idea. This mode made it really easy for me to play against Claude once it understood the game, then later allowed for running massively parallel simulations to test the various incarnations of the bots against each other.

None of this ground breaking… but I have never built a game and keep in mind we’re doing all this within the first four hours. 😱

It’s been super fun and was one of those projects where I hit same level of flow state that I used to get when coding directly, just at a higher level order. In the past I would have dreamt up backlog items for myself to create a CLI version, or to make a test harness for bot faceoffs. If I was lucky I’d get to those things days or weeks later if I got to them at all. In this case I had a tight enough implementation and testing loop that I never had to put anything my todo list, I simply implemented as I went and it was a completely joyful experience of just building whether I was typing the code or not.

Outside of architectural decisions like engine separation, and strictly sticking to client side only I STILL haven’t looked closely at the code (three weeks later). Maybe I’ll hit the limits of maintainability, or more likely the models will continue to improve and I’ll direct the required refactors, new test suites or repackaging required when I get there. We’ll see!