CuriousLM
I like the audacity here. Most AI startups are racing to build bigger data centers, and CuriousLM just puts the whole thing in your pocket and walks away. No account. That one line does more for privacy messaging than a hundred policy documents ever could.
The obvious tension is capability. On-device models have come a long way, but they still can't match what a datacenter-sized cluster can do. So the real question for me is what use cases they're targeting. If they nail the lightweight assistant stuff, summarization, quick answers, drafting, they could carve out a real niche. If they're trying to compete head-to-head with the frontier models, they're going to struggle.
What I'd want to know next is the hardware story. Does this run on a mid-range laptop or does it need a beefy GPU? Battery drain matters too. If they've solved the efficiency problem, this could be the privacy default for a lot of people. If not, it's a cool demo looking for a product. I'm genuinely curious where they land on that trade-off, because getting it right would be a real win.