Back on vacations and finally had time to come back to this again… so I opened my laptop, started Claude, and entered this prompt:
/goal Implemnent a backgammon game.
One player will be the computer
the second player will be a humam
computer will use AI to drive game play
all logic is implemented locally (no calls to external AI providers)
playing interface will be a browser based GUI
That’s all it took*. One goal. A fully functional web based backgammon game, with a nice UI and a computer player implemented in less than 200 lines of code. I did some small changes (meaning I prompted Claude to do the changes) to tweak the game play, but overall the results are mind blowing.
* it was actually a bit more than one goal – a review (done by Claude itself) after the first iteration found a couple of bugs… but for me the first iteration was already way above my expectations
The working game is here: Backgammon, hosted on an Azure Static Web App. The source code can be found here: vainolo/ai-backgammon-v5. I have NOT looked deeply at the code outside the “AI” player strategy, which is a simple heuristic, nothing fancy at all… which is also kind of mind blowing. But from a high level perspective it looks pretty good.
I’m very happy that I learned programming before coding agents because I got to build great stuff and really understand how things work “under the hood”. Software engineering is not dead, but it is now a very different thing from what it was just a 1 year ago. I’m both happy and sad because of this. I like building stuff “manually” but at the same time I do get angry at all the “manual” things we sometimes need to do to make things work.
Seems we are getting to the point were nothing is built “manually” and we can build anything we want, which is awesome. Because the biggest problem in software engineering is not building stuff, but thinking of stuff that is worth building. And now I can invest more on the thinking part.