Tailscale free: up to 6 users, each with unlimited devices (used to be 3 users)
I use Tailscale because it's professionally supported and better implemented. It has a much smoother UX compared to ZeroTier. It's also been profesionally audited, while ZeroTier has only sworn since 2020 that it'll totally be audited at some point after 2.0. ZeroTier may be a smaller target, but that also means it has less of a track record.
Also, I don't use Windows, but I don't see how that'd have anything to do with this.
It seems too centralized. At that point why not use a VPS? I have a free vps from oracle I already use to relay some unimportant services to a static IP. But it's extra latency
I also run monerod, p2pool, i2pd and tord nodes, the purpose of which is defeated by central infrastructure.
But yeah friends have reccomended tailscale I have just never gotten fuzzy feelings whenever I revisit it. If we want to build a NAT-less world then everyone needs to get a static ipv6 and run an eepsite or onion service, and we can just use pubkeys for addressing.
Not quite sure what your specific use case it if you've already got a VPS @ OCI (I do too, its great, and super low latency from home to it).
That said, another option not mentioned is cloudflare tunnels. I run cloudflared on a proxmox'd tiny PC. My internal services are accessible to me externally, but secure. Easy to set up the security layer (sign in through your google account, github account, IP restrictions, WARP and lots more).
I'm a long time network admin and Tailscale is my favorite network tool ever. It just magically lets you connect servers in a very secure way. The tools you mention are great for learning but miserable to actually use.
> the purpose of which is defeated by central infrastructure
Any ISP you use will also qualify as central infrastructure. You aren't making any sense. Just rent something that lives in a datacenter and get on with life.
> Make it void the warranty, don't allow customer support, whatever you want.
That would be a regression in terms of consumer rights, in my opinion. I don’t see why the EU shouldn’t legislate such that the right to control owned hardware becomes fundamental.
Technically, even in the US a warranty is only void if it's caused by something the consumer did; a 1 year warranty for production defects can't be voided.
Although I think it's fair that the company not have to deal with a defect that is almost certainly caused by something you do to your device. Like, strawman, but early iPhones sometimes encountered issues while doing an untethered jailbreak, or people would install tweaks that were less than stable (even sometimes on relatives' or friends' phones), then they would walk into an Apple Store with weird behavior and expect the fix to be free. That shouldn't be covered by warranty.
It 100% should be covered in the event the user is okay with the equivalent of a factory reset (they lose all their data). I agree Apple isn't on the hook for restoring the data you saved with a sketchy mod, but restoring the phone to factory functionality is 100% their problem.
I struggle to think of a reason why a phone can't be factory reset to fix those issues that doesn't involve "the manufacturer fucked around trying to prevent 3rd party changes". It's their problem and fault if their attempts to prevent 3rd party software ended up bricking the phone. I can't be bothered to give a shit that they've done weird things to the bootloader and now it won't boot because the bootloader doesn't have the right magic keys. Fix it or issue a refund, it's their software breaking the damn thing.
That does exclude "your software physically burnt out hardware" but that's such a rare occurrence it's practically not worth mentioning. I can't remember the last time I heard of a ROM that physically killed a part.
You're not likely to physically damage a computer by running your own software on it. There's no excuse for not having an easy "reset to factory settings" and a bootloader that's hard to brick. Anything with an SD card slot or a USB port can boot from it; anything with wifi can download a fresh factory image; anything at all can have a "recovery mode" that you don't need to touch to tinker with it. Tinkering is only dangerous when the manufacturer makes it so.
That’s well and good, but how are we supposed to evaluate the accuracy of random HN comments without anything resembling somewhat objective metrics? People say all manner of things, and usually it’s contradictory. What heuristic do you propose?
set up a rendering profile and preconditions that generates a minimal snippet of images/video using a predefined GPU profile.
then test for either a pixel perfect reproduction of the correct behaviour or for the properties you're looking for (if it doesnt reproduce deterministically).
this is one way. i also subscribe to the view that if the type system is modified to become stricter in such a way that it can fail reliably in the presence of this type of bug that this is also good enough.
some people might argue that these arent "strictly" TDD by some definition but they set out a path to follow red green refactor and confer identical benefits so my view is who gives a duck?
I don't have enough domain expertise to know which variant of these approaches is best but I'm enough of a TDD expert to know that what you're implying isnt possible is actually something you would would probably derive a lot of value from if you did it.
The typical playbook for a VC funded startup is to race to a monopoly where the company can have higher margins. Prices continuing to go down for the consumer over time would require competition to stay high in the long term, and even then it’s not clear if even current prices are profitable.
The current level of AI has plenty of inherent competition from local models. In the long term, most of the profit will probably be from very smart models that run at something closer to datacenter scale over long inference loops - where local inference can't do much and even third-party inference/small neoclouds will be severely challenged. That is a very natural "moat" and has natural cross-efficiencies with AI model training, which requires a similar scale.
I had that for a few months and cancelled. They have minutely rate limits as well so you get 3-4 hyperspeed responses and then a 45 second pause waiting for the throttling to let your next request through.
And then, depending on what you're working on, the 24M daily allotment is gone in under an hour. I regularly burned it in about 25 minutes of agent use.
I imagine if I had infinite budget to pay regular API rates on a high usage tier, it would be really quite good though.
> They have minutely rate limits as well so you get 3-4 hyperspeed responses and then a 45 second pause waiting for the throttling to let your next request through.
I haven’t really gotten that, though have noticed on some occasions:
A) high server load notifications, most commonly, can delay an answer by about 3-10 seconds
B) hangs, this happens quite rarely, not sure if a network issue or something on their side, but sometimes the submitted message just freezes (e.g. nothing happening in OpenCode), doesn’t seem deliberate because resubmitting immediately works, more often than not
> And then, depending on what you're working on, the 24M daily allotment is gone in under an hour. I regularly burned it in about 25 minutes of agent use.
That’s a lot of tokens, almost a million a minute! Since the context is about 128k, you’d be doing about 8 full context requests every minute for 25 minutes straight.
I can see something like that, but at that point it feels like the only thing that’d actually be helpful would be caching support on their end.
You must be on some pretty high tier subscriptions with the other providers to get the same performance!
reply