dovin

Just got access today. I like the channels, but I'm actually finding a few things frustrating compared to Dots: no phone calls (convenient to braindump and have it clean up later) and I don't have shell access to its underlying VM like with Dots. It's also weirdly hard / impossible to spawn Claude Code for Web workers and I can't connect the agent to a desktop and have it spawn jobs there. It doesn't seem all that faster than Dots.

It's early and obviously they'll keep iterating. It definitely feels less "make it yours" than Dots and that does feel intentional.

I use Muse to read X for me with a burner account, since I want to follow certain topics but don't want to use X. It posts digests a couple times a week to a discord channel I gave it access to. It's fine and I think there are plenty of ways of using it that give essentially nothing to Meta in the way of data. Still I wouldn't recommend it to any of my non-technical friends, since I think the natural progression with these things is to just connect more and more parts of your life to them.

Although, when there are a million Eunices running around trying to proactively help their users, they may have a harder time accessing that kind of data...

I wonder if this is going to create more market demand for truly open source models, where everything including the dataset is inspectable. It definitely seems like good news for companies like Mistral that want you to build on their models but they're not close to the frontier.

Can't decide if it's supposed to be ironic or not that this is an AI-generated article.

Jev does seem to have become the Kleenex of decision models. Is brand recognition worth $7.5B? There are lots of other decision models out there that perform at or near jev-level (laya, gliner 2.5 decide, even embedding gemma 2) that you can also run locally, and honestly I think this kind of model makes the most sense running locally as well. Maybe if TypeSafe can ship fast they can stay the default. Guess we'll find out.

Yeah, that seems like the kind of dataset they would like to keep in their back pockets

This does seem to be true in the age of throwing all data in to training the next iteration of the God Machine, but also, some data is a lot more valuable than other data and I want to know what the incentives are of people who are collecting data in our homes and what the actual data / derived data that they're after.

Yeah, it definitely makes sense to me if what adtech is often doing is just backing out from specific data to a general profile like income, location, etc, so that that level of targeting can work.

That's obviously bad and I hate it, but how much value even is there is the data that a spyware coffee machine could collect about your home? What advertisers would buy such data and what would they advertise to me? What is the marginal value of that data?

> As for Chollet, the whole purpose of the ARC-AGI benchmark is to determine true progress toward AGI. While LLMs are performing better on this especially tough benchmark, none have come close to hitting the mark.

I'm curious what "the mark" refers to, since it seems like every new iteration of ARC-AGI is getting saturated faster than the last one.

> If you genuinely, genuinely, believed artificial superintelligence was coming, why would you want to work at the company that would be in the best position to replace its employees?

I mean, there are a bunch of reasons one might. Making hugely impactful products can be great for your career, even if you only work there for some time. Maybe you believe you'll be better-positioned to see what's coming. Or cynically you want to make absolute bank so you can hide out in your New Zealand bunker when the AI dictatorship begins or whatever. Or you'd rather be a hammer than a nail, or in this case, the automator rather than the automated.

Or you might believe that superintelligence won't necessarily lead to huge job loss, won't happen that soon, or will bring about the fully automated luxury communist future so it's your moral imperative to be maximally accelerated.

Yeah open question what "perfect" search would even be, like would that just end up being indistinguishable from a kind of magical omniscience? And then there's "can I literally just find that one freaking slideshow from a while ago with that one client... or is it in Google Drive...?" And I really don't want the solution to be that we just plug everything into Claude