> Despite stated intentions, I think these platforms want to stay small and niche and non mainstream.
I think that's essentially it, whether they are doing it intentionally or not.
Running a social network is like curating a party or conference. Who you do and don't let participate, and who you do and don't promote above others, is essentially the product. And that is a lot of work.
Open protocols don't solve any of these problems, and people who care about open protocols don't tend to be the same people interested in curating giant parties and conferences for a mass audience.
You don't neccesarily have to fallback across models. Both Anthropic and AWS Bedrock provide the same models with the same API at different endpoints and different infra. So if you have a proxy endpoint to point users at, that specific fallback case is pretty easy.
I'm not saying that works in every business or billing scenario, though.
This is not super on topic, but ANW is the US version of the Japanese TV show Sasuke, which is itself inspired by the Adventure Zone segment on Takeshi's Castle. So, yes, it is connected.
> popular culture is much more fractured now into lots of small scenes
That's definitely a big part of it. But popular culture is also set up differently now where a different kind of person breaks through. There are some really great songwriters out there with Dylan-esque bodies of work (John Darnielle, Adrianne Lenker, Josh Ritter, Kristian Matsson are a few one might argue for). But the modern system isn't set up to create new Dylans.
A big part of Dylan's personality is his reclusiveness and mystery. He spun his own story, made up details, and generally just didn't reveal much of his true self in the press, but meanwhile the press constantly talked about him. But in the modern music industry, performers are asked to build an authentic social media presence before they can even get a record deal. If they want to go solo, they have to essentially build a direct marketing company. There's not much space for someone to be a mysterious recluse. It's part of the reason we have very, very few new "movie stars" in the classic sense of larger-than-life personalities.
They are actually selling really well. But sadly, they are selling because you can drill out the 'recording' indicator light and use them to secretly record people to make creepy TikToks.
Thats a good candidate for a closed fist falling at non-walking speeds at somebody's ugly face. Do that around my kids and you are automatically a pedo that will get adequate treatment.
Also, where I live its against the law to record people without their consent so if I would be feeling very nice I would just hold you against the ground till cops come.
You wear them = you film with them. Get your shit together kids, this ain't some harmless fun.
It seems to mostly just be a problem if you’re caught (aka if you’re being obnoxious or using a tripod) or if you try to use the subsequent footage for anything other than private viewing.
In many places you’re allowed to take wide shots where people might be captured incidentally, though not everywhere, as some places don’t allow photos of “copyrighted” buildings and landmarks.
I mean technically if you do it for no reason (it not actually recording), they have no video evidence, and if they did actually record and use it as evidence, well, it shows you were in the right. Not sure what punishment is for creeping on someone's kid vs getting a slap in the face tho
The article is cool, and it is fun to build a custom version - just sharing a little more info on this.
They've had this feature out of the box on a few Bambu models for quite a while now. It was part of the H2 series, and the A2L is them expanding it to the cheaper models.
The printers come with software to lay out text/drawings, line up the paper with the overhead camera, and run the job. And it comes with a tacky plate to hold the paper in place while drawing. It's all pretty slick - no need to mess with custom g-code or anything. You use the same process whether you are drawing, using a drag knife, or laser cutting.
But what I find works best is to point Claude at a design system documentation website (your own company's or another public source) and tell it to use that design style. It usually does OK, and the results are usually much more in line with that style and not as Claude-y.
This is such an interesting case because all sides are talking past each other and it doesn't really fit the pattern of past vendor lock-in / open source scandals. It's nuanced.
Empirically compared to most tech products in 2026, Bambu products have very little lock-in and are relatively open. But they are handling communication so poorly, they are creating their own drama and blowback for no gain while managing to piss off every OSS fan in the world.
Bambu's model is you can either (1) use the product off their cloud and do whatever you want OR (2) you can use their cloud but are locked to remote controlling the printer with their app. That's empirically more open than most home smart tech products (robo vacs, playstations, smart fridges, whatever) which don't give you the first option at all. Bambu literally lets you disable all their cloud and auth features with one click and use open source apps instead.
The issue is that Bambu made their closed cloud too good, so people really want to use it. Their cloud mobile app is way better then any other vendor.
Bambu has never done anything even remotely similar to HP requiring you to use their refills, yet people repeat this myth online like it's inevitable. Meanwhile the very popular eufyMake UV 2.5D printer literally has ink cartridges with lockout chips in them an nobody complains at all.
Where Bambu has totally failed is (1) they changed the cloud access rules after releasing their first products which people absolutely hate and (2) they are terrible at communication, making awkward threats at OSS developers who try to access their cloud without permission.
The funniest part is the only reason the developer was able to access their cloud is because Bambu supports Linux with OSS software. The Linux software isn't notarized, so they couldn't lock access to their cloud to a certificate. The 'hack' was just running the Linux cloud auth code on Mac/Windows.
This creates a situation where the actual details of the dispute are pretty nuanced, some people are vaguely mad at Bambu for things they absolutely aren't doing (like locking out 3rd party filament), and yet Bambu is still creating an army of haters to achieve basically nothing useful for themselves.
The smart move would have been to totally ignore the OSS developer and just lock down their cloud service on the server side. Then there would have been no drama and they would have achieved their goals.
Bambu needs to hire a western PR person who understands what is going on here to clean this up. The whole thing is dumb, avoidable, and pointless. Even if Bambu 'wins', they achieve nothing for the company.
The other nuance though is that Bambu is outselling other brands enormously and their sales this past Christmas were bonkers. No other vendor is even in the same league. HN readers will be buying Prusa or whatever, but Bambu is servicing a different market segment in which Prusa is totally irrelevant and has microscopic sales. But it's worth remembering that this entire drama is over something that literally does not affect 99% of Bambu's user base who happily use their cloud apps and have no reason to try anything else.
Essentially, you are adding another zero to the cost to have hidden solar. A 20k solar install becomes a 200k+ solar roof install.
Even if the final result is great, the economics shrink the possible customer base. Basic solar has gotten so cheap that people aren't worrying if the investment increases the value of the house itself. But very few people are willing to pay 10x for a thing that will never pay itself back in energy or home value. It's like putting a pool in your house - a few buyers will want it, but a lot will run from it because they don't know what to do with it.
So as a result, the target market ends up being super rich dudes in gated communities - the same kind of people buying custom 100k hifi systems and home cinema rooms. It becomes an upsell for people with unlimited budgets.
It's just not a mass market product when the competition is 10x cheaper and dropping daily.
Have you tried the latest DeepSeek v4 Pro inside of the Claude Code harness? It's not listed in that site.
It definitely 'feels like' it is as good as Claude for many regular web app coding tasks (though I don't have real benchmarks). And it is comically cheap.
I'm not suggesting it is better than the latest Claude or codex models, but it seems 'good enough' for a lot of use cases in my limited real world testing.
I'm starting to feel like a parrot, but people seem to forget that software engineering is actually a very narrow slice of the white collar pie. You don't need a mega-model which can reason about 100 000 lines of code when you want to create a nice PPT (which consumed literally hours of your life before) to impress your boss. SOTA models will probably be used for frontier research, complex coding tasks, large scale data analysis, etc. And the average Joe shall be able to buy a pre-configured box with a plug-and-play harness and run medium models air-gapped. Or use such models through cloud APIs dirt cheap if privacy is not a concern.
On the same topic but from a slightly different angle - as SOTA models get more capable, the 'quality' and 'feel' of the experience they provide in each domain is heavily dependent on the reinforcement learning the vendor does for that specific domain. After all, many fields have 100 flavors of "good answers," but the model has to pick one answer.
Benchmarks are not very good at capturing this yet. But it could be the case that DeepSeek v4 Pro is 100% as good as Claude Opus 4.7 at scaffolding a basic Rails app, but absolutely terrible at creating a credible business plan that another businessperson would think is real. That's a made-up example, but you get the point.
The end result will be a lot of people arguing about which model is "better," but "better" depends heavily on the task and how that model was trained to interact with the user for that task. Two users may have very different qualitative experiences using the exact same model, despite the benchmarks.
Creating a nice PPT is actually hard because it requires visual capabilities and so-called "computer use" (really, GUI use) of fiddly proprietary software. The nice thing about the coding case compared to a lot of disparate white-collar work is that it's all plain ASCII text. You can already ask a coding model to create a nice TeX/beamer slideshow (or whatever the Typst-based equivalent is) but whether your boss will be duly impressed by that is anyone's guess.
Tangential, but in our opinion corporate PPTX automation is an unsolved problem, even with Claude for PowerPoint (and it's worse with everything else common out there). Its harness (a) is not tuned very well for corporate use and (b) even if it were, fails to manage the specific business knowledge within each org needed to create effective (i.e. audience tailored) presentations.
I think that's essentially it, whether they are doing it intentionally or not.
Running a social network is like curating a party or conference. Who you do and don't let participate, and who you do and don't promote above others, is essentially the product. And that is a lot of work.
Open protocols don't solve any of these problems, and people who care about open protocols don't tend to be the same people interested in curating giant parties and conferences for a mass audience.