modeless

Angular resolution, not "screen real estate". It limits information density. Yes, the VR headsets have much higher FoV than a monitor, but it hardly matters for desktop work. Angular resolution is much higher on a monitor. Swiveling your head constantly to look at low-res windows floating all around absolutely sucks compared to having one monitor in front of you with higher information density. Varjo is the exception, and yes I'd say Varjo's resolution is good enough, but Vision Pro is borderline and Quest/Frame are much worse.

I have used most of the major headsets. Vision Pro is close to monitor resolution but not quite there. Quest 3 and Steam Frame are nowhere near that. Vision Pro's ergonomics are bad, as are Quest 3's, but Bigscreen Beyond 2 is much, much better and could be acceptable for laptop replacement I think, especially with the flip-up halo band they are working on. For inputs, passthrough with keyboard and mouse/trackpad would be fine for desktop work, and eye tracking could be used in combination I think, along with tracked controllers for 3D tasks.

Something with the ergonomics of Bigscreen Beyond 2, passthrough and resolution improved beyond Vision Pro, and the hackability of Steam Frame, I think could be pretty attractive as a laptop replacement for some people. Of course we're nowhere near that product yet. But a Steam Frame with a higher resolution screen and a better head mount (and the Arcturus Vision passthrough camera add-on) would be pretty close.

Never is a long time.

Given the current direction of technology it seems more likely that the need for highly precise and information dense UIs will be reduced, and we will do most of our work by directing agents through language. Where language is not enough we will use custom visualizations and purpose-built UI controls that contain and directly manipulate only the relevant information for the task at hand, constructed on-the-fly. We may not need highly general UIs based on abstractions and metaphors to the physical world anymore.

I think a hackable Linux PC in a VR headset is very exciting. But the major problem is the screen resolution is not good enough to make it a viable desktop replacement. It'll work great for gaming or media consumption, but reading smallish text will be difficult at best.

If they made a more expensive version with a better display it would potentially be interesting as a portable workstation/laptop alternative. But I don't think the SoC they're using supports displays much larger than this, so they'd have to upgrade that too. When you have such a wide FoV the number of pixels you need to cover it all at high resolution becomes very large.

All of the warnings about the danger of previous AI models were exaggerated, without exception. All of the proposals to coordinate nationally or globally are hopelessly naive. Everyone involved will soon regret handing control over AI to politicians.

In at least some varieties of strawberries, the top of the berry has the least flavor and sweetness, and the bottom has the most, so I don't feel like I'm losing much.

Taking a snapshot of the state of AI math right now and concluding that it will be net negative to human understanding and insight in the future is very short sighted. This statement will be used to promote ideas and actions that will ultimately be disastrous for our country.

I have 65 hours logged on Steam, am I allowed to say that I have also bounced off of several updates and find the game empty and lacking meaningful goals?

It's simultaneously an amazing achievement, worth the money I paid, a redemption story for Hello Games, and yet disappointing in a way that e.g. BOTW is not.

One of these years I plan to dive back in again, hoping to find that the sum of all the updates has provided some new story progression I will find meaningful and give me a reason to exercise all the new game mechanics. I'm hopeful!

> why house can be inherited indefinitely and creative work is not

Because copying houses is inherently expensive, while copying creative works is inherently free. To make creative works act like houses you have to add artificial cost to copying, which is a tax the public has to pay for the entire duration of copyright.

If the public has to pay a tax they should get something out of it, and that something is the incentive for authors to create new works. But does copyright have to be for life + 70 in order to incentivize new works? Clearly not! It could be much shorter while still providing plenty of incentive.

I'm not sure what your point is. Harry Potter is in the top ten highest grossing media franchises ever. Sure, top 10 franchises make a lot of money, but they're outliers. Is it in the public interest to grant every author in the country a copyright monopoly for life + 70 so that we can ensure JK Rowling's grandchildren continue to make billions from Harry Potter? Clearly not!

It encourages creation of new works just fine, but at the expense of creating billions of tiny monopolies that last for generations, which get aggregated into huge monopolies, ultimately making ordinary people into criminals for wanting to preserve or modify the works that make up their culture.

My point is that it doesn't need to do that. Copyright could still encourage creation of new works with a much shorter duration, without the downside of locking our culture up in monopolies for life + 70.

How about "We the people", as stated in the Constitution? The Constitution itself defines the purpose of copyright, and it is not the purpose you state.

Congress has currently chosen a duration of life + 70, but my point is that it's absurdly far beyond anything necessary to encourage the production of new works. When you consider the costs of copyright monopolies to the public, it's clearly out of whack. As I said, the majority of revenue from copyrighted works happens in the first couple of years of release.

The creator is not the only person to consider here. We don't grant copyrights just to give creators a funny kind of royalty-pension they can pass to their heirs. We grant copyrights to encourage production of new works for the benefit of the public. Copyright ultimately must be judged on how well it serves the public benefit, not on how much power or wealth it gives to authors directly (at the expense of others).

The major problem with copyright is it's too long. The vast majority of revenue from copyrighted content is made in the first couple of years, and yet we extend copyright 70 years past the author's death! If copyright lasted two years instead, all the problems with it would be much less severe.

Copyright is a monopoly that the government creates on purpose. We need to treat that monopoly power with more respect. We shouldn't be handing out generational monopolies willy-nilly. Copyright is allowed in the Constitution for "limited times" and Congress has ignored that.

It's hilarious that Meta made a good AI integration for Gmail before Google did.

For a while now I've had a startup idea for an agent that would monitor your inbox for content-free notification emails that just say "you have a new message, click here to read it" or "package delivered" with no details about the package contents. Then it would login to the website, do the 2FA dance, click through the UI, extract the actual important information, and put it back in my email inbox where it should have been in the first place.

I haven't worked on this idea because I think general purpose AI agents are likely to be able to do it very soon. And it seems like Muse is almost there already. It almost worked with MyChart, but hit a snag navigating the website after logging in. I expect this will be fully working in a matter of months.

I honestly can't think of any software ideas that won't be subsumed by general purpose agents or world models or some other form of AI in the very near future. Time to pivot to hardware ideas!

Does anyone else remember "Facebook M"?

In 2015 Facebook (not yet renamed to Meta) launched a personal assistant named simply "M" that lived in Messenger and functioned remarkably like ChatGPT does today. At the time it was basically entirely humans on the backend, with the idea that they could replace humans incrementally as they collected data. Really ahead of its time. They shut it down in 2018, which seems shortsighted in hindsight. That data would have been invaluable just a few years later.

https://www.technologyreview.com/2017/04/14/152563/facebooks...

So the timeline is:

Aug 28: OpenAI starts training a new model.

Sep 1: OpenAI sees a rumor on Twitter that two Millenium Prize problems were solved and starts their own effort to attack all the prize problems using the new (4 day old!) model.

Sep 3: The new model makes some progress toward Navier-Stokes. Based on this progress, OpenAI focuses on Navier-Stokes over the other Millenium Prize problems, using several approaches in parallel.

Sep 5: Navier-Stokes is solved. Assuming Astra API prices, $15m in output tokens were used by the whole effort.

In this account of the story, no specific information about Tristan and Levent's work is used to inform OpenAI's approach. The focus on Navier-Stokes and the choice of approaches to pursue came from OpenAI's own progress, not specific knowledge of Tristan's concurrent work.

There is a caveat that they "can't rule out" the possibility that Tristan's Codex data could have been part of the training set of the new model, though it is described as "unlikely" and the proofs are substantially different.

This timeline is insane. Navier-Stokes was solved start-to-finish in 5 days? A model in training for at most eight days dramatically outperforms Astra and Fable, and not just in mathematics?

> After hearing of the rumor, OpenAI started researching Navier Stokes with a new internal model.

I don't think this part is accurate. OpenAI was researching Navier Stokes before. It's possible that they started on a new approach after hearing of Tristan's success, however that is not proven and I expect we will hear OpenAI's side of the story today.

Because they didn't do that. Tristan doesn't specifically claim that they did, and Anthropic employees don't think they did either. https://x.com/_sholtodouglas/status/2097218240397410733

Apparently forcing is allowed in the Millenium Prize problem statement. So OpenAI's claimed proof could win the prize. OTOH the results Tristan and Levent are publishing here do not go far enough to win the prize, though apparently they are suggestive of a general approach that could produce a solution, which seems likely to be the general approach OpenAI's proof uses.

The question is whether OpenAI's pursuit of this direction happened spontaneously, or as a result of them learning about Tristan's work somehow. To be clear, while the tone of this post seems quite accusatory, Tristan does not claim to know for sure whether OpenAI unfairly benefited from his work. Sholto Douglas from Anthropic is also on record saying the suggestion that OpenAI used Tristan's codex transcripts somehow is extremely unlikely to be true[1], which I agree with, though it doesn't rule out them learning of Tristan's work some other way. I am sure OpenAI will have a statement out tomorrow clarifying their position.

[1] https://x.com/_sholtodouglas/status/2097218240397410733

Pretty sure you could unredact this based on the length of the name in this font. People who redact things don't really think very hard.

$360 per puzzle. When they tested people it took about 10 minutes per puzzle. If price/performance keeps falling at the same rate it has been, this will cost less than US minimum wage humans within two years. Three for Phillipines minimum wage.

It loses to Muse Spark 1.3? Does anyone really believe this index reflects reality?

> If I learned things at one company, took my experience and memory to another company, and applied what I had learned there, am I technically stealing that company's secrets?

The fact that California allows this to happen (banning non-competes and rejecting "inevitable disclosure") is exactly why Silicon Valley started here and remains here. It's exactly why ex-OpenAI people could start Anthropic. It's why neither OpenAI nor Anthropic have a monopoly on AI today. It's great both for employees and for the general public.

This seems like by far the best model yet for reconstructing 3D spaces from sparse images. It looks like you could reconstruct your whole house with pretty good fidelity from a dozen or so images taken on your phone.

They show it working with videos that have motion, but it seems like time is always frozen while the camera is moving, and they always return to a ground truth camera view before advancing time again. Maybe the temporal consistency isn't very good? This surprises me given how well it understands space. I guess modeling physics and time is the next step in the development of this kind of model.

More apt than you intended. Training wheels actually make it harder to learn how to ride a bike. I think this is what Carmack is implying about Scratch. I'm sure he would rate VB much higher than Scratch as an intro to programming.

It feels like an improved version of Celestia on the web, which I've wanted for a long time. The rendering is awesome!

I like the motion blur on the stars, but it increases the brightness of each star. It would look more consistent if the brightness of a star was constant regardless of motion. I guess the brightness increase makes things look cooler when you're warping through space but it's distracting when orbiting the camera around objects.

I think clicking to select things could be improved. Clicking on names is reliable but clicking on rendered objects seems unreliable. Sometimes clicking on a planet or star does nothing. Also my mousewheel seems quite slow to zoom, I know this is hard to tune across platforms on the web, I'm on Ubuntu 26 currently.

I can think of about a billion features to add. Real time clouds for Earth, a bunch more satellites and spacecraft, eclipses (Google Maps web 3D view renders these, crazily enough, if you set your system clock to the appropriate time), more terrain and high res imagery, multispectral imagery/rendering, relativistic effects, realistic close-up rendering of planetary rings, solar flares and fancier star rendering, binary stars orbiting each other, procedural generation of plausible stars and solar systems with exoplanets in other galaxies or unobserved parts of our own galaxy. Really depends on where you want to take it!

Personally I hate when I use a passkey but then still get hit with an SMS second factor step. A passkey should be enough, unless I'm changing my recovery email or withdrawing a million dollars or something. Also there's still a lot of really bad UX around passkeys, both by browser/OS vendors and by individual apps, and unimplemented features like sharing.

Passkeys are the right thing but they need more work.