lifty

Keep in mind that labs are now training their models for RSI so most likely the same training is useful for improving open source models, as long as guard rails are lax enough :) which most likely won’t last long.

Looks good but doesn't have a browser yes, like all the desktop applications of major harnesses have. Having a browser accessible to both the user and the model is extremely useful. I know you can do it via MCP but still, having a baked in one is better.

better start the dev guild or union so we can get some negotiation power in the political struggle.

you want to rewrite hof from scratch? Really curious what you're cooking, as I plan to adopt hof in a project.

Not sure what you're talking about. I strictly use the X website and it works great.

You can start with his appearance on one of the larger podcasts, like Lex Fridman (https://www.youtube.com/watch?v=Qp0rCU49lMs) because it will give you a good overview of his takes. If you don't like Lex Fridman just pick another big one, he's been going around. And after, just dive in his YouTube channel: https://www.youtube.com/@drmichaellevin/videos. He also has a podcast called Thoughtforms Life (https://thoughtforms-life.aipodcast.ing).

Rust + GPUI is becoming the new hotness, and I’m all up for it.

That there is another layer on top of the biological/genetic system which allows cells to organise themselves and pursue goals, taking advantage of bio-electricity (gap junctions in cells) in order to communicate. He uses the hardware vs software analogy which I think fits well. The theory sounds a bit wacky initially but all of it has been corroborated in empirical testing in his lab.

I am not a biologist but it seems to me, based on the podcasts that I hear, that his theories are becoming more accepted and there are also more and more people from outside the biology field looking into his theories of "diverse intelligence"; he has a podcast where he discusses related things with scientists from various fields.

Ever since I discovered Michael Levin and his work, I started seeing biological systems as information systems. And these papers seem to point in the same direction.

Aaand of course we can’t use it! GoOgLe iS bAcK iN tHe GaMe! There’s basically no way around it, all enterprises end up dysfunctionally shipping their org chart.

Random guy on X said that a couple of years ago. He’s supposedly an aerospace engineer.

There are scenarios where either can fail, that’s why it’s smart to have a multi pronged energy strategy. Maintain strong industries for being able to do both renewables and nuclear in house so you can hedge the tail risks of either of them. Not everything needs to have the most economically efficient model. As we’ve seen with the supply chain problems they happened during Covid, some things are more important than absolute cost.

I love renewables, especially solar with batteries, but I definitely see the tail risks of not having some amount of independent base load, for which nuclear seems like the best fit. The tail risk could materialize from different sources; could be an asteroid, nuclear war, weird weather patterns etc. nuclear would chug along in most of these conditions. So if I would be a policy maker or decision maker, I would definitely invest in nuclear and maintain an industry regardless of the initial cost. I would try to make it as cheap as possible.

Finally got the nuance. Indeed the chat part of the subscription is unlimited as far as I know. Now that part of your comment makes sense!

That’s the default. I didn’t express myself clearly but I thinking your situation is not the common case anymore, or perhaps you are not using it hard enough. Codex limits deplete very fast these days, it’s not “unlimited”.

It’s all anecdotal based on my experience and other countless discussions I have seen online. I’ve heard speculation that once they hit 20 million codex users capacity is tighter so they have to manage it. The previous limits were unsustainable compared to token pricing.

What are you talking about? ChatGPT unmetered? No way! That was 2 months ago perhaps and it’s possible your account still hasn’t gotten the new limits. I noticed around 1 month ago I was still going full throttle on my codex subscription and my limits were barely budging, and then all of sudden people around me started to complain about limits. I thought they’re crazy, but then my account go the hammer, and that was it. If I have the same pattern of usage like I did before, basically having an agent working continuously on a coding take, my weekly limit goes in 2 days.

Can you buy these already? I think not.

I’m having the same issue. Hold max subscriptions on both frontier labs but I’ve been forced to use open source models because token limits are not what they used to be. So I end up using Astra and Fable for reviewing, and open source models for implementing.

Is there any chance bcachefs will be included again in the future?

I was wondering, do you think its possible to use something like SAM 3 (segment anything from FB) + Laya to create a super efficient and fast computer use tool?

I think at one point you said you’re working on a rust rewrite? Is that still in the cards?

Meant as a joke in good spirit. I do appreciate what Google is doing but I am rooting for them to also compete on the frontier.

Google is crushing it! They’re going toe to toe against the European giant Mistral with all these small useful models and they have a chance of winning! Never have I been more excited about Google’s capabilities.

That TPU training part I get, but from what I have seen the NPU is rarely used for inference by LLMs. They still use the GPU, no?

Do current models run on the NPU or GPU? Wondering if Apple will have something like a TPU.

Unfortunately the codex plans don’t offer the same amount of tokens as they did before. This changed around a week ago. There’s been a lot of user reports noticing this issue, and I’ve noticed the same pattern on my account. Previously I would never reach my weekly quota but last week I managed to finish it it one day. Same project, same single session sequential work. Not sure if there’s an issue or if it’s on purpose, and not even sure it applied to all accounts. Curious if other users on HN noticed the same problem.

Great primitive, I’m a big fan of prolly trees! You mentioned git like filesystem snapshots? How would you retrofit that on an existing filesystem? Or you’re talking about writing a new one?

Thanks for the detailed reponse!

I’ve been keeping an eye on hypen.space and hope it succeeds. What is the level of maturity now and are there any multiplatform app examples that are not trivial?