We implement STUN and TURN functionality natively in WireGuard rather than using separate protocols.
Netrinos uses a central rendezvous server that participates in WireGuard handshakes solely to collect your devices' public endpoints and share that information with your other devices. When a device roams to a new location, the server learns the new endpoint and updates the other devices in your account.
When direct P2P fails, Netrinos connections fall back to a relay server. The relay is a WireGuard peer, but it can only relay traffic between peers in your account. All customer accounts are strictly firewalled from each other.
If you want more control, you can enable a device in your account as a relay server with a checkbox in the app. This could be a home PC with a stable connection or a low-cost cloud server.
Thanks for the feedback. The document model in mongo was slopped together by a junior engineer, so perhaps an unorthodox approach. It is basically flat, and already used in a pseudo-relational manner via in-app join to the existing sqlite store. This blog post inspired me to think, what if we just chucked all the json from mongo into sqlite and used the generated indices? Then we can gradually "strangler fig" endpoint by endpoint
In my neighborhood (a Criminal Justice Reform Zone), the catch and release of repeat criminals caused a surge in break ins. The citizens organized and funded the installation of Flock LPRs. Several criminals have been caught as a result, and crime is now down.
So the impetus is twofold:
- Funding provided by programs such as Operation Stonegarden and other grants
- Activists agitate for Criminal Justice Reform --> Surge in crime --> The People clamor for Enhanced Security Measures and DIY
> the catch and release of repeat criminals caused a surge in break ins
> Several criminals have been caught
The actual difference here is that the second "caught" isn't followed by "and released". The camera didn't do it.
My street has repeat offenders who come and steal from cars nightly. The cops know who they are and have arrested them multiple times, with them immediately being released AFAIK. A million cameras wouldn't change this.
The community got together, worked on a solution, that solution lead to arrests. A politically savvy prosecutor would not easily dismiss an organized community with proven ability to drive results.
Cameras with good software work great for that, however the data should NOT be freely accessible outside of the city/jurisdiction they surveil. That's the issue with Flock vs any other AI camera/database product.
There is a trend towards less personal accountability and more centralized prevention. Instead of properly dealing with people who misuse sharp knives, we are making all knives duller.
The city I used to live in trialed flock cameras for car theft. They caught more car thefts in January of the trial year than the previous year’s total.
We started hoping that car thefts would be a pressure point for a lot of violent crime (which tends to be committed from stolen cars --- this is the Kia problem). But we caught more innocent drivers with stale entries on the Illinois LEADS hotlist than actual stolen cars. When we OK'd the system after its pilot, it was on the condition that we no longer curb cars based on stolen car reports at all --- we'd only curb them based on stolen license plates (which have no innocent explanation).
Maybe other states are different for this, but in Chicagoland, unless you don't care about disproportionately harming Black motorists, using Flock for stolen car enforcement was a flop.
The lesson I keep getting from your experiences is that LEADS needs an overhaul.
It turns out other states do have flags for things like "extraditable warrant" vs. just failure to appear warrants (something mentioned in previous discussions), and perhaps something could be done about the LEADS system if attention was given to it. It seems like fixing one's data sources is a great approach vs. tossing the baby out with the bathwater — unless of course that's the intention all along, as it is with many opposed to state-owned surveillance of this nature.
This is not exactly an unbiased forum to discuss this matter since Flock is a YC backed program, but what do you think will happen in short order? Maybe that car thieves will simply slap on fake license plates to get out of the area?
What you’re left with then, is nothing but the tyrannical and even treasonous mass surveillance program to know where you go and when all your life, even when you leave your tracking device phone behind and use a tracking device free vehicle.
Nobody cares that Flock is a YC company. I'd be surprised if most YC batch members even realized off the top of their head that Flock is YC. YC companies get criticized all the time on HN, including by people who have done YC.
But suddenly adding cameras that resulted in catching more people fixed the issue? Surely if the catch and release was the issue, that wouldn't make a difference.
Of course, you should instead believe your own anecdotal evidence that marks a tally every time you hear about a crime, which if you watch local news is always, but doesn't mark a tally every time there isn't a crime.
Fun fact that's totally not related, did you know
that people who listen to true crime podcasts are more likely to believe a crime could happen to their families and are also more likely to install higher levels of security on their house?
psychopath Scam Altman does not give a rat's behind about your "privacy"; he is merely trying to keep the grift going and avoid responsibility for his unethical behavior (see also: Scarlett Johanssen's voice)
So first off, no shit the college board things the SAT measures good stuff, it's their test.
Second nobody said the SATs don't measure something, but that something is ability to take an SAT test which is highly predictive of how well you can take other tests. Which as our society puts lots of stock into tests isn't nothing but it's not measuring anything inate.
China cut fossil fuel use 2% in H1 2025 (compared to H1 2024). It cut coal use 2% in H1 2025 (compared to H1 2024) [1]. The majority of new energy production is wind and solar, not nuclear, and certainly not coal.
As of past 1.5 years, the use of fossil fuels for electricity generation in China seems to have reached a plateau. They are building new coal plants, more new plants than they are retiring old plants. But even with new coal plants, they have not been burning more coal.
They are barely building any nuclear plants, so far zero finished in 2025.
In terms of coal its usage has started to decrease in actual terms since the start of 2025. All demand growth and more is solved by new built renewables and storage.
How does your relay compare to Tailscale's (DERP)?