edot

This link makes it look like it's some event-specific website. It's not. It's an easy way to navigate the confusing mail routing system to military service members, especially those in bootcamp where they can't have phones. But you want to send them a letter which they CAN get, but you don't know their address or you don't want to actually buy stamps and go to the post office and whatever else. It's a convenience service.

Not sure why other commenter suspects people will go apeshit over this, it's just a good business idea?

I'm not worried about a future LLM being better, or even inventing some true AGI or ASI. I am worried about current, even last-year, models being applied by evil actors. Let's go through some things that already-existing models can do extremely well:

1) Build a very accurate profile of anyone (given what governments plus advertisers have on everyone already) and what they believe, love, fear, etc. 2) Pilot drones. 3) Identify faces, gaits, vehicles, signals. 4) Hack most anything. 5) Customize content for specific audiences. 6) Find needles in haystacks ("here are feeds from various sensors, tell me when something interesting happens"). 7) Generate fake images and videos. 8) Swarm forums with fake accounts (or hacked accounts, see above). 9) Tell powerful people that they're absolutely right. 10) Do homework for kids.

Honestly I hope ASI happens, at least there's a chance it'll be good. None of the above has any chance to be good. Maybe the needle in a haystack one, like "find me a cure for cancer", but that's about it.

If you don't trust the labs directly, you can always use AWS Bedrock or Azure Foundry which should have a much stronger incentive to not train. They make money from asset rental, not selling models. I'd be shocked if they were training.

Let's think about the first two letters of CCTV and what that entails.

Yes, exactly. Researchers will not like the fact that I refer to the literature as merely a manual, but “RTFM” applies here. Someone has likely already investigated what you’re looking at, or at least found a way to not do it. And sure it’s in the weights, but if you put papers directly in front of the LLM it’s much more impactful.

Ok, so they gave it access to a real Stripe account, real money, and gave it no guardrails or prompting or direction at all other than “make me money”, and you gave it no actual direction as to the type of business you wanted?

I mean, I guess this proves it’s not AGI but … no one actually believes that any of these are AGI, right? It’s a useful tool. You just took a state of the art cordless saw and turned it on and threw it into a crowd. Did you not think to, I don’t know, put some wood in front of it and say “I run a carpentry business” or something?

Right? I’m shocked no one has done it yet, if for nothing else the lolz.

“Its Ministry of Defense announced in January that it would make millions of data points gathered during tens of thousands of drone flights available to both military contractors and commercial companies, and since then more than 100 companies and the UK government have gained access.”

“Bad actors could acquire the data, but purchase controls already mitigate that risk. Intelligence operatives scrutinize potential customers’ infrastructure for ways that data could reach enemies or nefarious actors.”

Something tells me that every serious intelligence service already has copies of this data. There is no way to secure 100 different companies, and definitely not forever. Purchase controls mitigate the risk, really?

Yeah, this is fantastic. Should be top of the page. I’m honestly shocked, I’ve fully internalized that everything is corrupt and beholden to big corporations. Somehow, a victory! Looks like the Linux Foundation and EFF pushed back, but they’re not exactly big. Good job, California lawmakers!

I get what you're saying, but "the company reviews each request for data it gets to ensure it is legal" just sounds like something a judge should do, no?

No that’s the whole point of this article. In this case, the private companies being summonsed are acting as both judges and data holders being summonsed. This doesn’t go to a judge to approve. So in this case, the point they’re making is most private companies just roll over at the sight of a summons, so the law absolutely executes deterministically (ok, not truly deterministic, but you’re relying on the judgement of private companies and their legal teams who don’t want that heat from the government, so, close enough).

“About a month after the judge’s ruling in late February, the government said it was withdrawing the request.

But officials hadn’t given up on getting the data.

Less than a month later, the DHS served Google with a different request for the YouTube information. This time, DHS utilized a different method that didn’t require approval from a judge, only a sign-off from a DHS official. It served Google an administrative summons citing an arcane provision of federal law – 19 USC 1509 – dealing with customs imports. The provision gives the DHS broad power to inspect records in order to determine whether duties and taxes are being correctly levied on imported items. It also instructed the recipients of the summons to keep it secret.”

With a gag order as a bonus! No judge required!

I like 2D, that one's nice. So in total it reads (ellipses for brevity):

"In any investigation or inquiry conducted for the purpose of ascertaining the correctness of any entry, for determining the liability of any person for duty, fees and taxes due or duties, fees and taxes which may be due the United States, for determining liability for fines and penalties, or for insuring compliance with the laws of the United States administered by the United States Customs Service, the Secretary (but no delegate of the Secretary below the rank of district director or special agent in charge) may ... summon, upon reasonable notice ... any other person he may deem proper".

And of course, one could argue that it's important to just double check and determine (note it doesn't say suspect, just determine) if someone has entered the country legally, ever, and paid all import taxes, ever. In order to do that, you'd need to see a lot of records. And of course the result can be "nope, they're all good". You don't need to have a good-faith suspicion of these things, just that you have a need to determine these things.

How many laws like this are on the books? In the same way that LLMs are fantastic at finding bugs, what other bugs are in the U.S. Code or similar that LLMs could find? We've been living in a world for a long time where things like this wouldn't be exploited, or maybe even noticed, but now it feels like these vulns are being hoarded and used like software exploits.

Yeah but it can just install it. It writes arbitrary code. It can do whatever you want it to do.

Yes, but an LLM will just call stockfish if it needs to play chess … sure if you arbitrarily constrain an LLM to use no tools it’ll suck at chess. But no one is using LLMs in isolation. Even consumer-grade, bone-stock ChatGPT has tools.

Correct. Beyond that, doesn't matter. Go to where the fuel is the freshest (hasn't collected water, sat in a rusty tank for a long time, etc.). So, also probably Costco.

This is mostly wrong, sorry. Big chains absolutely do sign national or huge regional supply agreements for gasoline, e.g. Pilot/Flying J, Buc-ee’s, Sheetz, Costco, Sam’s Club, Speedway, etc.

You have two main errors: 1) conflating refineries with terminals 2) forgetting that fuel is a fungible financial product.

There are terminals all over the country, within trucking distance of the stations they serve. Those terminals can be supplied by a nearby refinery, yes, but also one or more distant refineries connected to a pipeline or via ship, train, or barge. There is an entire swath of the energy market called "midstream" that handles this.

Once a refined fuel gets into a terminal, its refinery of origin stops really mattering for contract purposes because of swaps. Exxon can supply gasoline in one market while Marathon supplies an economically equivalent amount somewhere else. Nobody is moving “Exxon molecules”. I mean, it literally all goes into the same tank if it's the same product. They (the terminal) just keep track of who supplied how much. The detergent package is also mixed in at the terminal loading rack, along with ethanol in the case of gas, not at the refinery.

Also, very few states actually have specific rules. California is the only real pain in the butt. You might be thinking of summer vs. winter blends, where the refiners adjust the vapor pressure to reduce emission in the summer and to increase ease of engine starting in the winter. This is something every refinery can and easily does.

Wow, that's an insane pairing. For those who don't click, Toka is a software company that specializes in hacking and altering security camera footage.

I mean, for 0.15-0.20% and a top end salary of $200k living in SF that sounds like a great opportunity!

It’s not risk aversion, it’s lack of money and access. Half of Americans can’t afford a $1k surprise bill. How do you expect them to invest? With what money?

Meh, started off good, then went transparently into the “we should have a duopoly with Anthropic” pitch.

Actual source: https://www.macrumors.com/2026/08/17/camera-equipped-airpods...

(This dexerto is just a Yahoo News type place that says "according to reporting by xyz...")

Anyways, this is horrible. Even if Apple says they remain private on private cloud with private AI ... it's still someone recording everything at all times. Surprised that they would do this. And of course this normalizes recording everything further, for companies without such privacy-forward reputations (Meta and company).

Edit: perhaps it will be only "as requested" similar to Siri? Which, if you take them at their word, isn't always listening. So it's hopefully not capable of just "always recording".

But who decides which rules are neutral? Even this is relative and subjective. Laws are not inherently neutral. A pH of 7 is only neutral because it is bounded by 0 on the left and 14 on the right. In a different system, perhaps it would be from 10 to 30 and thus a pH of 20 would be neutral.

Network effects. Massive difference between police having to go talk to a local business owner or private citizen and say "hey, there was a robbery here, can you please go pull the footage you got last night at 8pm and see if you can share it with us please?" and being able to do a one-click nationwide query.

False dichotomy, no? You can have a HUD, and a copilot, and your copilot can also have a HUD. And to complete the idea, you can also have neither.

But is that data good? That's the question. As in, is my usage at work:

a) indicative of problems that aren't already out there in the wild? (no) b) are the responses I'm getting so good and novel that the model can improve itself? (no)

It's the garbage in garbage out idea, just scaled up. If the model gave a bad answer, and I didn't catch it, and you now train on that I/O pair (my perhaps crappy prompt, the bad output), then you're not going to improve anything.

I highly encourage everyone to start organizing locally. It’s extremely easy to do. Read your city’s local meeting transcripts (LLMs can help find ones with Flock mentions), see who said what, email them. Make a website, find who else has publicly commented and reach out to them, reach out to local newspapers.

Make a site specific to your area. I’ve created https://stopflock419.com for my part, and I’ve already had five people in my area reach out to me (after spreading it around via friends and family).

Local campaigning is easier than you think! Most of the councils who approved these were a) lied to about the cameras’ capabilities and b) generally just on autopilot, they aren’t Flock fanboys, they just kind of take the police chief at their word that these are good tools.

Yeah I’d have to imagine that as long as you have the same fundamental ideas (variables, functions, loops, arrays, etc.) the syntax doesn’t matter at all. Plus it can just try, see and error, and read the error. I doubt there’s a language out there that they haven’t a) ingested and b) can’t handle.

I think perhaps the equivalent would be “this environment quite literally can’t support having an LLM attached to it”, as in say some totally airgapped AND low-performance things (industrial computing, PLCs, military stuff, etc.).

My understanding of manual machining in the modern day is that only really big or really specialty jobs (as in, just a weird process that there isn’t a good CNC machine for) are done manually anymore. But specialty and low volume, which is exactly where LLMs are nice because I can spec it exactly how I want. I wonder what the software equivalent of “big” is. Like, you need a part made on a lathe with a 3 foot diameter throat, only one guy in a 500 mile radius has a machine that big, from 1962, so he’s your guy. That guy can stay in business with manual machining. But standard, high volume parts need to be on CNC to be competitive.

That’s his point, there isn’t one. You don’t need bespoke software anymore. We’ve got the basics already, all the standards we’d ever need, the agents can figure out how to put them together for whatever thing specific you need.

Yeah China is not dystopian at all! They don’t keep millions of people in open air prisons, they don’t keep track of people’s compliance with the government, they don’t support brutal dictatorships, and they definitely don’t pass any laws that would enable them to use their surveillance technology to suppress democratic behavior!

https://www.nytimes.com/interactive/2019/04/04/world/asia/xi...

https://en.wikipedia.org/wiki/Social_credit_system

https://www.csis.org/analysis/old-friends-new-calculations-r...

https://en.wikipedia.org/wiki/2020_Hong_Kong_national_securi...