growt

If the tech is mostly in the glasses temples, I would prefer to have something that attaches to my existing glasses. I have prescription glasses and a big head, tried the Even G2 smart glasses and the available frames set just too small for me

Can you give a link to your paper(s) here? Sounds really interesting.

So smol is your agent? Do you have a GitHub link?

Wishlisted! Looks great

I wouldn't mind if similar _common_ words are mixed up. But the current state of the iOS keyboard and also gBoard is that it sometimes just makes really unhinged suggestions (at least in my native german). Words I never used before and that I didn't even know existed. I think a small AI layer on top would probably fix that, just the likelihood of next words multiplied somehow with what I possible could have swiped.

Where and how do you run that? I tried it but somehow I always ran out of context or generation was incredibly slow (mbp m4 pro 48gb).

Even more fitting is the part of the story where a collective of uploaded lobster minds are involved. I wonder if that was an inspiration for the "OpenClaw" name somehow or just pure coincidence.

There is a similar experiment where a famous violinist plays in a subway station. Nobody really notices or appreciates him and his music. https://m.youtube.com/watch?v=hnOPu0_YWhw&ra=m

I just did a quick check: a 2022 Macbook pro M1 with 1tb and 16gb ram goes for 1119 euros on eBay. A new M5 mbp with the same ram and disk is 1899 euros, so 709 difference. And you get a whole new machine not just a cpu upgrade.

Well you can sell your old MacBook. And they hold their value pretty well. So I don’t know if framework would actually come out ahead financially.

Wasn’t cache time reduced to 5 minutes? Or is that just some users interpretation of the bug?

1) bully or bullys insurance 2) whoever sat on it Alternatively: Apple care? :)

That is just spec driven development without a spec, starting with the plan step instead.

The main difference could be that you have an existing code base (probably quite extensive and a bit legacy?). If the llm can start from scratch it will write code “in its own way”, that it can probably grasp and extend better than what is already there. I even have the impression that Claude can struggle with code that GPT-5 wrote sometimes.

An interesting fact about the early GTA games is that they owe their success to a bug. The cop cars were supposed to behave nicely like in every other game, but due to a bug in pathfinding they just drive straight into the players car. So at least to some extend the whole billion dollar franchise owes its success to a bug: https://medium.com/@bdunn313/the-psycho-cop-bug-de9121335cf9

After reading through the other comments about bending spoons and reading yours again: the bending spoons CEO is technically telling the truth! They intend to run the acquired companies forever. After cutting most of the staff, but he didn’t say that part of course.

I don't have any links but you can start with CLAUDE.md and/or AGENTS.md and put the basic instructions in there ( you can also google these filenames for examples and recommendations). I also put README.md's in every subfolder to describe which file does what, etc.

It might be 1), being an early adopter doesn’t help much with AI. So much is changing constantly. If you put a good description of your architecture and coding guidelines in the right .md files and work on your prompts the output should be much better. In the other hand your project being legacy code probably also doesn’t help.

Probably depends on where you are, but here in Europe I always joke that you should prepare for the exact opposite of what apple weather tells you. A lot of times I’m literary standing in the rain and Apple tells me the chance for rain is 0%

From my experience as a parent, that market is also very small, because the time between “child is old enough to text and be away from parents for long enough” to “child wants to have a real phone” is not that long.

If you’re aiming for the second phone market you don’t have to beat the iPhone. Probably the easier pitch.

I recently build a small framework to create JavaScript apps that use this kind of URL sharing and therefore don’t require a backend: https://github.com/grothkopp/lost.js

I think this is most likely what happened. The update/review process for extensions is broken. Apparently you can add any malicious functionality after you’re in and also keep any badges and recommendations.

Lost.js – Local, Offline, Shareable Tools

Lost.js is a lightweight, zero-dependency framework for building shareable "vibe coding" apps and prototypes.

A week ago my wife asked me for a "wheel-of-fortune"-style app as a fun way to choose random climbing routes. I thought that's something I could "vibe-code" myself in a few minutes. But then, as always, comes the user-management, db setup, backend development, etc. So this time I thought I'll do it differently and the result is lost.js: a framework (or toolkit, or maybe just template? I can't decide) that is small, has zero dependencies and no need for any backend. The data/objects are stored in local storage in the users browser. But they are shareable, via special URLs containing the whole data in compressed form in the URL hash. The lost-apps are all PWAs by design and work offline once they're cached.

So if you want to code a small tool, that lives in a browser lost.js can be your starting point and allow you to focus on the tool itself without worrying about backends or user-management or sharing.

I flashed it. Don’t know if always on works with the browser. But jailbreaking is easy and I think even reversible. Then you can ssh into it and even setup a cron to download a picture and refresh the screen.