jowday

Worked adjacent to the AV space 5~ years ago. This wasn’t my area but I remember learning that this was a robustly solved problem long ago.

GoreTex is a bad example - it's gonna delaminate after a year or so of heavy use and is pretty much impossible to repair after that. Which also undercuts ACRONYM's messaging about their GoreTex products being some kind of like, buy-it-for-life rainjacket.

I’m sure there’ll be a PS6, but it honestly seems like non-portable consoles are on their way out. Wouldn’t be surprised if there’s no next-gen TV console XBox at all, or if it’s just a MS branded gaming PC.

From someone that worked in the industry ~6 years ago, it's clearly going well for them - frankly, they're expanding and scaling way faster than I would have thought possible in 2019. They've got something like 6 cities running right now and what, 3-4 more announced?

Another thing to keep in mind is that rideshare revenue in the US is extremely geographically concentrated in urban cores. This is why every AV company was targeting SF as their first city (excepting Waymo, which did some stuff in PHX). 'Hyperfocused expansion' probably looks a lot closer to tackling new, novel areas in different metro areas rather than, say, expanding down in to San Jose and the central valley.

These things, they take time.

They've clearly hit (or projections confidently show they'll hit) a point where each car is profitable. I worked in the space for a while - platform upgrades (new cars, sensors, etc) are planned out years in advance and are pretty complex processes. But generally, each upgrade was a massive decrease in cost per car. (usually 50% cheaper or more). So also possible they want to wait for the next platform transition.

Weird, I thought it was one of the best movies I've seen in the last few years. Wasn't at all what I expected to see, but was incredibly memorable and impactful.

F1 on the other hand was maybe the worst offender as far as literalism is concerned.

Dude, I explicitly replied that I pinpointed it to Next. The moment I removed the Image tag and replaced it with a regular <img />, it resolved itself. I didn’t even say I was using three, did I?

The most confusing part is that it only happened really intensely on just one developer's machine.

I'll admit that the way we were using the image tag was a little unusual, but still something that was imminently supported by a plain HTML image tag.

My point is more that Next is such a bizarre black box that things like this were a regular occurrence.

To clarify: yes, it was the next Image tag. The moment we switched to using a plain image tag it resolved itself.

Lee Robinson from Vercel is gonna show up in this thread as soon as he wakes up, don’t worry.

Don’t ever use Next. Terrible developer experience, vendor lock in, weird undocumented conventions that make building anything other than some kind of B2B SaaS CRUD site full of undocumented foot guns. My favorite thing I’ve encountered is the Next <Image /> tag somehow dropping the FPS on a webgl scene on the same page to 2 FPS.

> Plus, you've got the very large challenge of learning a rich, high-quality 3D representation from a very small pool of 3D data. The volume of 3D data is just so small, compared to the volumes generative models really need to begin to shine.

Isn’t the entire aim of world models (at least, in this particular case) to learn a very high quality 3D representation from 2D video data? My point is if that you manage to train a navigable world model for a particular location, that model has managed to fit a very high quality 3D representation of that location. There’s lots of research dealing with NERFs that demonstrate how you can extract these 3D scenes as meshes once a model has managed to fit it. (NERFs are another great example of learning a high quality 3D representation from sparse 2D data.)

>That said, our belief is that model-imagined experiences are going to become a totally new form of storytelling, and that these experiences might not be free to be as weird and whacky as they could because of heuristics or limitations in existing 3D engines. This is our focus, and why the model is video-in and video-out.

There’s a lot of focus in the material on your site about the models learning physics by training on real world video - wouldn’t that imply that you’re trying to converge on a physically accurate world model? I imagine that would make weirdness and wackiness rather difficult

> To be clear, we don't yet know what shape these new experiences will take. I'm hoping we can avoid an awkward initial phase where these experiences resemble traditional game mechanics too much (although we have much to learn from them), and just fast-forward to enabling totally new experiences that just aren't feasible with existing technologies and budgets. Let's see!

I see! Do you have any ideas about the kinds of experiences that you would want to see or experience personally? For me it’s hard to imagine anything that substantially deviates from navigating and interacting with a 3D engine, especially given it seems like you want your world models to converge to be physically realistic. Maybe you could prompt it to warp to another scene?

Why are you going all in on world models instead of basing everything on top of a 3D engine that could be manipulated / rendered with separate models? If a world model was truly managing to model a manifold of a 3D scene, it should be pretty easy to extract a mesh or SDF from it and drop that into an engine where you could then impose more concrete rules or sanity check the output of the model. Then you could actually model player movement inside of the 3D engine instead of trying to train the world model to accept any kind of player input you might want to do now or in the future.

Additionally, curious about what exactly the difference between the new mode of storytelling you’re describing and something like a crpg or visual novel is - is your hope that you can just bake absolutely everything into the world model instead of having to implement systems for dialogue/camera controls/rendering/everything else that’s difficult about working with a 3D engine?

The paper they’re basing this off already does this.

https://diamond-wm.github.io/

It’s essentially this paper but applied to a bunch of video recordings of a bunch of different real world locations instead of counter strike maps. Each channel is just changing the location.

https://diamond-wm.github.io/

This is pretty much the same thing as those models that baked dust2 into a diffusion model then used the last few frames as context to continue generating - same failure modes and everything.

https://diamond-wm.github.io/

What was hard to understand about it?

There are immersive theater-type events that can scratch the same itch, although they tend to focus much less on participation than LARP events like this. The Starcruiser is probably much closer to poorly-executed immersive theater than live action roleplaying. Really well done immersive theater can be incredible - Sleep No More in NYC is one of the most mindblowing experiences I've ever had (although it closed earlier this year). Legitimately felt like I was stepping into a David Lynch movie every time I went. It only had limited, somewhat improvised audience participation with the actors.

In my experience, the best immersive theater experiences find very clever ways to make the atmosphere work. In Sleep No More and other Punchdrunk shows, all of the guests are given masquerade masks to wear, the venue is fogged, and the lighting is dim. The dim and foggy atmosphere hides stuff that would otherwise take you out of the dreamlike 1920s noir setting of the show - that the other guest walking next to you is wearing a graphic tee, for example. The masks cast the audience as a shuffling horde of vengeful spirits haunting the characters for the sins they commit throughout the show - so when you see a big crowd following a character, it doesn't instantly feel at odds with the setting the way I imagine seeing a 5 year old in a Pokemon t shirt on the Starcruiser would.

Next has been a nightmare to use every step of the way. Endless undocumented ways to shoot yourself in the foot, especially if you’re trying to deploy somewhere other than Vercel. I can’t imagine anyone using it for anything other than the most basic CRUD app. And even then I’d recommend not using it.

I commented on your other thread - but you should really clean up your information diet. From the vocabulary you use and the tone of your game and pitch deck, I can tell you’re someone that hangs on the words of tech industry ‘thought leaders’. You need to realize that most of the people in tech who have time to podcast, write substacks, or otherwise build a ‘personal brand’, aren’t actually making shit. They’re trying to inflate their profile so they can trade reputation for career advancement in any number of ways. It’s also not worth listening to most VCs. Most of them don’t have the time or technical ability to understand the areas they’re investing in and just chase trends. If they had the time and technical abilities, they’d be building companies instead of getting other people to do it for them. You’d be surprised to hear how many investors or personalities that are supposedly high profile are openly derided among actual founders.

The way to make something that’s fun is to try to make something that’s fun over and over again until you’ve got it down. It’s not by obsessively reading what investors or people who are essentially glorified influencers say.

I’ve evaluated almost any model that’s come out for game asset generation in the last few years and none of them are up to snuff. That alone tells me you haven’t looked at the output of these models closely or aren’t familiar with Gamedev.

Also, I’m sorry dude, but your blog sounds exactly like every metaverse pitch I’ve ever seen. Down to the “ROBLOX, MINECRAFT, ETC” highlight at the end. This doesn’t make me feel like I’m reading something written by someone that cares about games - it makes me think the author reads a lot of VC substacks and played League of Legends for a while.

Even mentioning “GTM strategy” and “ideal customer profile” tells me you’re drinking a lot of Kool aid. Stop watching YC videos aimed at B2B SaaS founders and reading a16z blogposts and work a lot harder on this stuff if you ever want to show it publicly.

Also keep in mind, there’s probably literally 100 other companies with the same pitch, vibe, and idea that you have.

I feel like you’re trying to launch a consumer product like it’s a B2B SaaS company. As it is (broken site with no polish, little to no gameplay features, product hunt banner) you seem like you’re to trying to chase AI/metaverse trends. Reminds me of all of the endless metaverse scams and scandals. I think you’re significantly underestimating the time and polish it takes to build a consumer 3D web based product . Most small game studios take years to build up even small games before releasing them.

If I were you, I’d think deeply about whether this is actually something you’re passionate about working on or if you just want to start a tech startup and chase trends. If the former is true, I’d close this down and work on it for a lot longer before ever showing it publicly. If it’s the latter I’d pivot to an LLM B2B SaaS company like everyone else like you and try your luck at that.

That’s common parlance in archeology for ancient languages in that region.

https://en.m.wikipedia.org/wiki/Ancient_Semitic-speaking_peo...

As someone that experienced it several times, the red ring of death was a 270 degree circular segment - a full red ring indicates cable failure.

I think most rage bait about the Warhammer series is manufactured by content farm channels to generate clicks - the series wasn’t even officially in development until a few weeks ago.

There’s an entire subgenre of YouTube channels that consist solely of creators updating videos promising that they have inside information on the creative conflicts at Lucasfilm/Amazon/etc, all of which happen to align perfectly with whatever the fandom is outraged with that week.

https://youtube.com/@mikezeroh

This guys channel is a great example - most of the channels discussing “female custodes Henry cavil warhammer 40k tv series” Amazon follow a similar format.

Edit:

I’d also add that I don’t think Rings of Power is bad because they cast minorities - most of the actors are fine, really. The plotting and pacing is just horrendous. In a show that has 5+ active plot lines and threads scattered all over the world, they’ve spent a quarter of their screen time on a plot line that’s completely disconnected not just from the lore but from the wider story being told and doesn’t look like it’s going to connect anytime soon. Which is funny, because I’d imagine Amazon execs felt that they were obligated to include that plot line (the Hobbit one) to appease viewers.

It’s not really consistent - or anymore consistent than, say, SDXL with IP adapter. Even in their example images the character they’ve input comes out wearing different clothes.

Same way reverse engineer get access to internals of other devices - a bunch of tricks. :^)

In one case, greentheonly realized some fraction of Tesla’s cars are shipped out of the factory still in dev mode, with debug mode enabled and increased privileges. He found someone with a car like this who was down to helped and swapped part of his cars hardware with their car, and from then on was able to get a much better view of what was running on his car.

Unfortunately twitter is awful to search and a lot of his info is buried deep in old threads, but a few (old) examples to illustrate that he regularly does this.

Visualizing the outputs of the models running in the car: https://x.com/greentheonly/status/1404164587927314435

Tesla’s dev debug menus circa 2020: https://x.com/greentheonly/status/1336467014727110656

It’s been a while since I’ve worked in the space so I haven’t followed green as closely.

I’d advise you to ignore lay explanations of the space - outside of the industry most of the discourse about self driving cars is poisoned by Elon’s deceptive presentations and his followers who parrot what he says.

If you want a grounded explanation of how Tesla’s stack works, follow @greentheonly on twitter. He’s a Tesla reverse engineer who regularly posts about the software that’s actually running on the car.

If you want an explanation about how real AV companies stacks work, I’d read Sebastian Thrun’s robotics textbook - then imagine what’s outlined in that book but with ML plugged in to a ton of spaces throughout the stack. This is also similar to how Tesla’s stack works, btw - greens just good to follow because a lot of people refuse to believe Tesla isn’t running some kind of “LLM but for driving” fully end to end black box model.

Tesla won’t launch a robotaxi anytime soon because they can’t use remote support or HD maps - although I think they’ve been stepping up their mapping efforts. Even the demo at Universal studios a few weeks ago was HD mapped - per @greentheonlys twitter.

I worked in the space for years and have seen the internal of both a traditional robotaxi company’s stack and Tesla’s.

Neither of them are “full AI/ML”, they’re both traditional robotics systems with ML used for detection/prediction/planning at certain steps. Elon will sometimes say something about moving to a “new ML stack”, but Tesla reverse engineers regularly look inside of what’s running in the cars and that’s not the case at all.

Contrary to what other people in this thread are saying, the remote support isn’t remote direct driving of the car - essentially what will happen is that if the car finds itself in a situation where it’s unsure of how to proceed and it’s safe to stop, it will pause for a few seconds and wait for a remote operator to clarify a situation for it.

A good example might be road construction - if the car detects new road construction work that doesn’t match its map of the area, and its onboard systems determine that it’s not sure how to proceed through the construction with confidence, it will send what it thinks the top five likeliest ways to proceed to a remote operator. The operator then selects the proper path (or says that none of them are proper). The car will then follow the path presented by the operator, but actual driving behavior /collision detection / pathfinding is still determined locally. Think of it like ordering a unit around in StarCraft.

Fundamentally different - people will flock to LiDAR vs Vision when debating this, but the more fundamental difference is that Waymo uses HD mapping + remote support.

Usually the people that give information to outlets in cases like this are directly involved in the stories in question and are hoping to gain some advantage by releasing the information. So maybe this is just a tactic that’s not as favored by Anthropic leaderships/their counterparties when negotiating.

The first thing Waymo tried building (way back when, circa 2010 or so) was highway based driver assist in the style of Autopilot. They did a ton of testing with it and didn’t like how quickly their testers stopped paying attention despite a ton of instruction not to. I’ve seen clips from these tests.

It’s also possible they shifted direction because the long term vision of robotaxis is much more lucrative.