sensitiveCal

“Smart brevity” works well for updates, but I’m not sure it translates to thinking.

When everything is compressed into bullet points, you often lose the intermediate reasoning which is where most of the actual insight lives.

At some point it stops being a writing style and starts shaping how ideas themselves are formed.

Feels like this is treating AI as something to be excluded rather than something students need to learn how to work with.

Typewriters might force focus and remove shortcuts, but they also remove iteration, editing, and research — which are core parts of modern thinking and writing.

I wonder if a better approach is teaching students how to use AI critically (and verify output), instead of designing environments where it simply isn’t available.

The interesting shift here is framing this less as an addiction problem and more as a coordination problem.

It’s not just that people can’t leave, it’s that leaving only works if enough of your network leaves with you. Otherwise you’re the one who loses access to relationships, context, and communication channels.

Feels similar to other network effect systems where the “cost of exit” ends up being higher than the cost of staying, even if the experience itself is worse.

Feels like this is sitting somewhere between Ollama and something like LM Studio, but with a stronger focus on being a unified “runtime” rather than just model serving.

The interesting part to me isn’t just local inference, but how much orchestration it’s trying to handle (text, image, audio, etc). That’s usually where things get messy when running models locally.

Curious how much of this is actually abstraction vs just bundling multiple tools together. Also wondering if the AMD/NPU optimizations end up making it less portable compared to something like Ollama in practice.