What kind of “value-generating” features are you thinking of?
One big comment I want to give on that front is, if you decide to try any generative AI features, please please please make them *completely opt-in* and integrate with *local models* (either completely or as an alternative to cloud models like GPT / Claude / etc). If I can use your software happily without being bombarded with ads for AI features and can use those AI features how I want to instead of how would best make you money, I would appreciate that so much.
You got it. Yes, AI will be a good piece of the puzzle. Local models is exactly what I prefer, if you haven't notice, that's a big principle of this app. I'm really hoping Apple can step in and offer a local model so I don't have to download one big blob. Supporting Ollama sounds cool but I feel like that's a big barrier for many... BYOK also nice, but at that point why not just build a MCP interface instead of building another flavor of AI chat? Otherwise I'm thinking about embedding all items so it's easier to: auto-generate tags, cluster by topics etc...
Comments
What kind of “value-generating” features are you thinking of?
One big comment I want to give on that front is, if you decide to try any generative AI features, please please please make them *completely opt-in* and integrate with *local models* (either completely or as an alternative to cloud models like GPT / Claude / etc). If I can use your software happily without being bombarded with ads for AI features and can use those AI features how I want to instead of how would best make you money, I would appreciate that so much.
You got it. Yes, AI will be a good piece of the puzzle. Local models is exactly what I prefer, if you haven't notice, that's a big principle of this app. I'm really hoping Apple can step in and offer a local model so I don't have to download one big blob. Supporting Ollama sounds cool but I feel like that's a big barrier for many... BYOK also nice, but at that point why not just build a MCP interface instead of building another flavor of AI chat? Otherwise I'm thinking about embedding all items so it's easier to: auto-generate tags, cluster by topics etc...
as initial step, just do some “rag” / vector db thing ie. no llm. see chrome’s embedded model and how it works.