Non-deterministic nature of these models is definitely a concern. Interesting to see how companies implementing generative interface circumvent this. Maybe they create a set of fixed generic UI elements or write a system prompt describing general design guidelines. Still wouldn't fix it 100%.
I love ink & switch! I think Martin, Geoffrey, and the whole team are crazy smart.
There's a whole bunch of approaches to universal versioning and I think floro is one of a few upcoming techniques in the ether.
From my limited understanding of their approach, it looks like they're going to be really focused on things like rich text diffing, which is not something I think floro is particularly well suited for. I think floro is really great for things that end up getting consumed as code rather than something like note-taking. e.g. static assets & i18n.
It's really exciting to see this space actually beginning to blossom though!
I think the importance of this study is being downplayed by many.
That same spontaneous and "breakthrough" nature of the abilities of LLM that is now nullified, was a major component of many arguments against LLMs and could've possibly stopped us from achieving greater intelligence.
I see this as the removal of a possible great filter event of the evolution of LLMs. This is a pretty big concern being nullified.
>Human body is an inelegant and unoptimized vehicle for intelligence.
Why do you think so? the impermanence of it? the gradual performance decrease? the biological/physical limitations that constraint it?
I don't think it ever failed in the sense it never shipped because that was out of scope.
It was never meant to be an end-product. It was a concept/case-study. Happy they're finally realising that vision.
