Every guide for solo devs says the same thing: finish what you started before you start something new. This week I started two games.
The last few weeks were heavy. A hard month at home, a new job behind the wheel, applications going out, and 16 guides written in seven days. I hit the point where opening my own codebase felt like lifting something. So I did the thing that sounds irresponsible. I opened Unity.
Scaling Laws
An AI company tycoon that starts on January 1st, 2022, with 12 million dollars. And this one got out of hand in the best way.
- The model quality runs on the actual Chinchilla scaling law, roughly twenty tokens per parameter.
- There are 22 real accelerator generations with their real ship dates and launch prices. A B200 bought in January 2025 is worth about 53 percent of what you paid, the same day.
- You can rent compute at cloud prices lagging 180 days behind, or own hardware at a third of the cost per FLOP and eat the depreciation yourself.
Which makes the whole game one sentence: you are not buying upgrades, you are timing them.
127 tests pass, including one that plays a scripted four-year campaign and checks whether the player survived and stayed competitive. No art yet. No playable build yet. Both repos are public, from the first commit, so you can watch the whole thing be wrong before it is right.
And yes, one of your rivals in Scaling Laws is a certain AI lab that starts quiet and then does not stay quiet. I will let you guess.
Baka Bake Bakery
A bakery you watch grow from a food truck into a wooden shop. 14 baked goods, a manager called Mila who takes over when you stop clicking, a five-minute day cycle. The rule I set: the world is the interface, and automation has to stay worth watching.
This week
- PC Workman rests. Deliberately. It has had a brutal month and so have I.
- Scaling Laws gets its first ugly playable build. Grey boxes, no art, just enough to feel whether the economy is fun.
- Taxi keeps funding the move.
I used to think rest was what you did after finishing. Turns out it is sometimes what lets you finish at all.