site banner

Tinker Tuesday for October 13th, 2026

This thread is for anyone working on personal projects to share their progress, and hold themselves somewhat accountable to a group of peers.

Post your project, your progress from last week, and what you hope to accomplish this week.

If you want to be pinged with a reminder asking about your project, let me know, and I'll harass you each week until you cancel the service.

1
Jump in the discussion.

No email address required.

Still learning more about having a better workflow and using AI agents better to get better results and have a cheaper token to work done ratio. I will not be using Python in the future, as I haven't had to go in and edit any code myself, and instead will have the AI agents code in Rust instead. I've been so impressed with the output I've expanded the scope of the project and what the app can do.

I asked the AI for a rough estimate of how much it would've cost a human developer to build out what I have so far.


Size: about 28,400 lines of Python, not counting blank lines and comments.

  • Core (map reader, MPQ archives, minimap, sprites, triggers, translation): 32 files, 11,300 lines
  • GUI (PySide6 windows and dialogs): 35 files, 15,100 lines
  • Tests: 7 files, 2,000 lines
  • Docs (README, notes, todo): 8 files, 1,400 lines of prose

The git history covers 231 commits. The first commit was about 1,300 lines, so nearly all of this was built in six days.

Time for a human developer: roughly 7–14 months full-time (about 1,200–2,500 hours) for one experienced Python developer working solo.

  • A productive developer averages roughly 75–150 finished, tested lines per day over a whole project, counting design, debugging and rework. At 28,400 lines, that's 190–380 working days.
  • Adjusted down: AI-written code tends to be wordier than what a human would write for the same features.
  • Adjusted up: much of the project is format research, not typing. Working out undocumented font formats, matching the game's text layout from screenshots, iscript sprite rendering, and handling protected maps' decoy data can each eat weeks, and few developers already know Brood War internals.
  • For comparison, the standard COCOMO model gives about 80 person-months for this size. Too high for a solo project, but the right order of magnitude.

Cost to hire it out:

  • US freelancer at $100–150/hr: roughly $120,000–$375,000
  • Overseas freelancer at $30–60/hr: roughly $40,000–$150,000
  • Agency: often 1.5–2× US freelance rates

These are ballpark figures. Lines of code is a rough proxy for effort, and the real cost would depend heavily on finding someone who already knew the StarCraft map formats. Without that, the research is where the hours go.

You also did real work that doesn’t show up as lines of code. You defined the features, tested every stage on Windows, kept a disciplined commit-per-step workflow, and kept the project docs current. On a contract, that would be the product owner’s and QA’s time.


AI estimate so taking it with a grain of salt but I find it incredible the value AI can provide now. Most of the people here are probably already well aware of this but I guess I never really understood what AI can do until I used it myself to build something out.

What this means is that anyone can now build hyper specific tools specific to you for something that could've costs hundreds of thousands of dollars just 5 or so years ago. Since this is a tool hyper specialized for me, I likely was able to skip a lot in the design and testing process that would've added much more time, but this just means it's very viable to just have AI build out tools for you. I guess the future is now.