2001zhaozhao

I think there is something here with the DAG exploration graph that systematically increase the surface area of exploration when the agents QA's an app.

But i suppose that this is only useful for codebase-wide bug sweeps? If one actually integrates it into a feature-by-feature agent loop, then the QA following each change is already narrow (since it only needs to test 1 changed feature), so this kind of exploration thoroughness optimization (in author's own words "recall") is less important.

> It's far, far, far more invasive, omnipresent, and required to play (like Fortnite)

Did they do something to Fortnite that I don't know about? I thought the only thing Fortnite did was allow pay-to-win in their community games (just like Roblox), otherwise the main game is still cosmetic only like Valve's games.

The criticism is right in a vacuum, but suppose that Epic Games got their way and successfully replaced Steam. Then what? We'd just end up with a new corporate landlord over PC gaming, albeit one that is also a private company and maybe charges lower fees, but has somewhat less proven reputation than Steam to not make future greedy changes to needlessly screw over users.

Astra = Agent-4

The dreaded cloud egress tax, except now it's actually a tax.

If an alien came to my backyard in an advanced spaceship then spoke English like GPT-2, I'd declare that we have met a new advanced form of alien intelligence too.

Touch on Mac OS is pretty cool news. It's not that useful in the current form factor, but hopefuly iPads running MacOS is next on the list of their products (i.e. basically an Apple version of Surface Pro).

On a side note, I hope this makes Linux distros follow suit with good touchscreen support. On a lot of them Firefox webpages don't even scroll properly on first install.

Hmm, fair point, I suppose I just don't use the hand tracking stuff very often since it's just not very productive (I use portable mouse & keyboard even on my Samsung headset that supports hand tracking).

But even given that the glasses have to support hand tracking and 6DOF, I think they can probably build the sensors into the glasses but compute them with the phone. Or maybe calculating hand tracking would drain the battery of a Pixel too fast or something. Not sure.

I actually don't see any reason why I would use this instead of my Xreal Air (released like 3 years ago) + Samsung Dex. The phone is a perfect "compute puck" for wired AR glasses and Google even makes them, so why the heck are they selling separate compute pucks with these glasses and making them unaffordable?? Just make some pixel glasses that run android xr on the pixel phone instead

Wordpress is quite unnecessary now, but so are other CMS's as you can just use AI to make a CMS with exactly the features you need from scratch. AI can also point you to existing frameworks that completely replace Wordpress functionality like page builders and then help you deploy them to the cloud, often as static sites that cost literally zero money to host.

I still have a WP site that I've used for the last ~7-8 years and I am planning to migrate off it in the near future.

This has been really cool to see and i have a lot of game/mod ideas that are borne out of the implications of this.

AI has been absolutely great for serious game dev as well. I am now completing dozens of tickets a week on my game engine and building features and architectures I've planned and theorized for years but never had the time to actually build them. Seeing them come to fruition one by one over the last few weeks and months has been incredibly rewarding.

For example, I've long wanted to do a specific architectural approach that allows my Minecraft server's players and standalone game players to crossplay with each other. Now it's actually happening!

I think the problem in the future will not be making games, it will be standing out because games are becoming easier and easier to make nowadays! There are still many game genres, especially multiplayer, that suffer from not having a big enough playerbase and too much fragmentation can definitely worsen the game experience there.

This is ingenious, lol. I might steal this for deterring video game cheaters?

I'm a Discord server admin and I'd love to turn this on because having more discussion of my game indexable by search engines seems like a good thing in general. For new messages it seems like there's no problem, in combination with the opt-out option. But for existing users, how am I supposed to disclose the change to people? What if people posted content in the past and wouldn't want it on Google, but aren't active anymore?

Furthermore, what if someone wants to opt-out of their messages being indexed in some Discord servers but not others? I don't see that option being available right now.

I totally get the task board and I do almost the same thing myself. The demo idea is cool though. It makes sense that whenever the coding agent wants some judgement from you it spends a bit more tokens to make a report optimized for a complete and readable explanation. I doubt it should be a video (videos impose a minimum time required to get through them), but there's certainly a lot more you can do compared to coding agents' default text responses.

I like how there's a giant "chapter picker" pinned at the top of the web page, but there are only two chapters and the second chapter is a tiny blob at the end of the article.

Very cool. I think a focus on AI code review and continuous knowledge graph synthesis are both right directions, and any system that provides the infrastructure to automatically handle these tasks eventually must do the job of a Git Forge, even if it runs locally.

I am tackling the problem from a different direction, but I think there's a lot here to be inspired by.

I was a subscriber of their now-unavailable coding plan. The 1000 TPS was very unique at the time, but nowadays their qwen 3.8 gets roflstomped by Mercury 3 at both cost and performance at similar TPS speed. There's not much reason to use them unless you're an enterprise with a custom deal.

I think getting a good reputation for being able to build real, ambitious software with AI will be a really important thing going forward.

This comment made me think that someone should pull an ArtCraft on Android or iOS by making a good linux phone skin, with high performance auto-updating wrappers for various web apps to make them as convenient as mobile native ones. Then we can actually ditch the big corps and bask in our LLM-enabled open source utopia. Hmmmmm.

I am glad that at least one Chinese company seems to be listening to AI cybersecurity concerns and did concrete steps that actually reduce the problem. This was probably at least somewhat costly to them in terms of general model capabilities as well. They'd probably have gotten slightly better coding capabilities if not for removing the cyber training data and tasks.

The problem with GA is that it's just so damned inefficient.

I considered using it on codebases as an outer loop to make LLM vibecoding-optimization loops more reliable, but imho it's a lost cause because the massive compute you need to simulating a useful population size can easily be spent on something else instead, such as just running more agents to enforce better engineering patterns and fix bugs in the first place.

I have liked diffusion models for quite a while, ever since i used the original Mercury to play AI Roguelite. I only have one question regarding Mercury 3: how tf did they get it to be so cheap?

If these benchmarks hold up it is a no-brainer against Haiku 5.5 and GPT-6 Luna, especially considering that Haiku 5.5 becomes 5x more expensive past 100k context, and this model goes up to 1M for the same price. (Note: the price increases 2x after a period post launch but is still way lower than Haiku is priced for >100k requests)

PLEASE NO

answer: you make a recurrent neuralese model

(btw don't do that)

I only skimmed through this post. But I think this is a crux:

> The more we read [the neural network] the way its own readers do, the more interpretable I think it will look. We should be wary of reading it too much like a human, and concluding it’s messy just because it doesn’t fit our preconceived concepts.

The problem with advertising "vibe coding" is that anyone can do it.

I think any personal AI agent at this point is "very sticky", not just Muse. You really aren't going to switch when a lot of the value is in the agent memory and there's no easy way to export that, similarly to how you aren't going to switch away from Spotify when it's learned all your song preferences which you can't export. Right now, maybe the agent memory is small enough that if you ask the agent to export you'll get most of it out (Claude has notably made official export instructions to encourage switching from ChatGPT in the past), but in the future the memory will be too large and complex to do that effectively, and platforms could also block the memory export or degrade it deliberately.

So, in the long run you pretty much have to pick one personal agent and stick with it, and switching is costly (unless you are using an open agent that lets you just download the memory, in which case it actually becomes cheap, because AI agents make other parts of switching way easier in general). For the purposes of Meta this means whatever user they get they'll probably keep.

HOWEVER, the thing that Muse doesn't have is network effects. They're running into the exact same problem as ChatGPT in this area, IMO. There's no reason for you to be running specifically Muse even if all of your friends are running it. So that gives other vendors a chance to catch up.

Cloud gaming is a big inspiration for some of my own innovations indeed. But why run expensive server-side GPUs and stream video over the Internet when you don't have to, and can still get the same anti-cheat properties? :)