Cloud AI is incredibly capable, but not everything belongs in the cloud.

I started experimenting with local AI to see how much could be done on hardware I already owned — while keeping personal documents, photographs and other private information at home.

What began as an experiment has grown into a working private AI lab — and an ongoing lesson in what local AI does well, what it doesn't, and how much tinkering sometimes hides behind the words “it works.” 😁 

The system continues to evolve, but today I use it to experiment with local AI chat and reasoning, image understanding, private knowledge, and image generation.

One of the next challenges is semantic photo search — being able to search a large personal photo collection by describing what I'm looking for rather than knowing the filename or folder.

Local AI is capable, but cloud AI is still better at many tasks. Integration and tooling are often harder than model inference, privacy comes with tradeoffs, and I've learned that getting something working isn't necessarily the same as making it useful.