> When deep space exploration ramps up, it'll be the corporations that name everything, the IBM Stellar Sphere, the Microsoft Galaxy, Planet Starbucks.
You've got to be careful, though. Consider that they have non-deletable apps, tiny amounts of onboard storage, and that every "app" WILL bloat over time. So, with those updates installing the latest version of Netflix, Prime, HBO Max, Paramount, Tubi, LG Channels™ etc... it means it's filling your 4GB or whatever storage fuller and fuller. Eventually the storage will be nearly full, which shouldn't matter, except that it's a computer, and having storage nearly full makes it slow as shit to do everything, even "boot up" and "change inputs".
> Google Maps (at least in the US, in my experience) is really great about telling you what lane to be in.
Having recently driven in Boston[1], the problem is that Google Maps tells you what lane you need to be in when the route divergence happens... it doesn't tell you what lane you need to be in to get to that lane.
So I would see guidance like be in the first or second lane (from the left) to make this left turn and all seemed well since I was in the second lane from the left. Then two more lanes merged in from the left and I had to get over two lanes and ended up on a two mile onramp for the highway to New Hampshire.
I see the same things locally, but it's not as obviously bad because I'm familiar with the local highways and also most of the local highways aren't as space constrained so there's almost always plenty of time to merge.
[1] I don't recommend it... we should have driven to the airport and returned our car and taken transit back to our hotel
This information is all derived from overhead (satellite and aerial) imagery combined with street level imagery (like from Google street view). No road owners required.
Where in the world are these road owners, and do they really eve provide this kind of information to mapping companies?
Here in the States, most of the roads are "owned" by the people and our various thousands of governmental jurisdictions do not tell anyone about these kinds of details in any uniform way.
That's also where this map would help, it can definitely degrade gracefully. If I know I have two right-hand turns coming, I can take the decision for myself.
> Today, having a piece of furniture in our home depends on a massive production and supply chain; the "average" person no longer has the ability to build it themselves.
Tangential, but this is one of many reasons that I (and I suspect many others here) have taken up wood working.
I tried this out this week and liked it but really wish this project had some form of auth. Opening the contents of every page you’ve ever visited, even to the local network, is not the best idea.
Hister supports token based, password based, and OIDC/OAuth authentications with optional multi-user handling. Details about user handling can be found here: https://hister.org/docs/user-handling
It also has a "public mode" where anyone can search the indexed content, but only authenticated users can add or modify it.
It seemed like the public mode was the default when I set it up. If so, that’s a fairly dangerous default as keeping a “clean” history with no secrets leaked seems neigh impossible.
The default configuration binds only to localhost, and a fresh installation starts with an empty database/index. Could you clarify which specific attack surface you are concerned about in that scenario?
I’m not concerned about an attack scenario. I’m just saying that using the docker image, if someone (or their agent) isn’t careful, they could expose their browsing history publicly fairly easily. It might just be nice to default to at least a user and pass login rather than just wide open.
Including your AI agents. And models become problematic. There very likely is and/or will be lots of pressure for US AI labs to make models help law enforcement. It could be by implementing backdoors in generated code, or not report some exploitable bugs, or something else. Similar for agents, they basically become the threat in your infra…
(It’s of course not only the US, just that the largest AI providers are US based and we know from history how US agencies operate)
I'm using open weights models on a privacy-focused provider right now, and that's adequate for my current usage, but I'm rapidly getting to the point where my agent's access level to my data (and to a lesser extent, my accounts) is becoming something I'm not comfortable sending outside my network at all.
My hope is that models that are roughly on par with Deepseek V4 Flash can be run on hardware that I can own for <~$5k in the near future. We're close, but not there yet as far as I know.
The only long-term solution to this is self-hosting.
There is literally no protection or difference of an opensource model doing this.
An actual objection I had while talking to an aerospace company was they don't want opensource models because the threat of it having a poisoned training example on specific systems.
It's easily the most hidden malware possible, completely undetectable until an exact set of tokens unlocks it. Is it line 100,543 of your security product? You will literally not know until it plants it in there.
To the best of my knowledge (which, admittedly, is far from comprehensive), that kind of attack on a model shouldn't actually be possible in a deterministic fashion.
If you can't stop them from sometimes telling customers things like "yes, I will give you a penthouse suite at our hotel for only $3/night", why would you be able to guarantee that, with some specific set of tokens, they would produce a perfect and undetectable backdoor customized to the code at hand?
I don’t think it has to be perfect or deterministic that way. It’s enough to have a bias towards implementing a backdoor in some circumstances. Something like, if the machine seems to be used in a Chinese environment, the model is biased towards missing some security issues, or towards implementing the type of bugs that can be used for an RCE, or similar.
You can review generated code manually or with models from a second vendor (ideally from a different country). Attackers would have to poison both and do it in a way that also makes them ignore the planted malware when reading code.
Open models also let you read reasoning traces. That means anomalies would show up when the backdoor activates, like a run of unrelated words or a jump in top token probability. It's only undetectable until the first time it happens.
