@aaacat
For AI CLI programming tools, the biggest advantage I’ve felt is being able to run them directly on servers.
You set up proper tests, define environment constraints, manage everything with Git, and run it in the background with tmux.
After each run, the AI validates whether the results meet expectations—or whether there’s room for optimization—and keeps iterating.
In the future, product iteration may be measured in hours or even minutes.
The backend will need many agents executing tasks in parallel.
When everyone is using similar foundation models,
being down for even an hour can put you significantly behind.
At the same time, simply adding more compute isn’t enough—you need matching use cases.
Otherwise, extra compute is just wasted capacity.
0 reply
0 recast
0 reaction