It sounds like you said it is simultaneously far left and far right?

This is so weird. I haven't visited the site the early Afghan War, maybe early Iraq War era, but I remember it being rather US-jingoistic, pretty right wing? And the site tilt was pretty heavily "white American male" not a famously left-wing demographic.

Meh, I knew of Fark back when all that was at the site was a single photo of a squirrel with large genitals.

I used to play on the mud Three Kingdoms in the mid '90s, and Drew was a Wizard (admin) there (named "Cletus"). When Fark became a link aggregator, I used to visit for the Photoshop contests (then ended up drifting off to Worth1000 when that became a thing). I'm not surprised that he ended up launching a porn site, because, well, Drew, but they launched a TV service that ended up getting gobbled up by Adult Swim? Weird.

Crazy to see all of the stuff that's happened with the site since I left; it entirely dropped off my radar.

> I think this is the result of their trainers trying to reduce hallucination.

Most probably. I don't like LLMs burning my precious tokens blathering on about being honest, but I will accept the tradeoff. Grudgingly haha.

> My iron rule is that Claude is not allowed to modify git state, not even to stage or unstage, and definitely not to commit. As long as that rule is followed, any mistakes can be cleaned up easily.

I have a repo in a sandbox with no remote. Claude is quite good at pushing commits to this repo during its own milestones during development, and it will use it to hunt down bugs when tests fail after a commit, stepping back to a commit before this session's work to see if this is a new bug , or a new way to expose an existing bug. If it sees that it is a new way to exercise an existing bug, it will binary search its way through the commits until it narrows down what changed, and then looks at how to fix it.

I disagree with all your assertions.

> AI has access to your command line to autonomously execute commands. No no no.

Ok, I must admit that I did cast that line off the back of the good ship ColaMan without any further information in order to see if anyone would bite.

You think that particular command line has anything of worth in it? The only thing in there that I can't recreate with one command is the local Git repository that the LLM commits its changes to. And that's only because I manually copy it out of the LLM's sandbox on occasion. I should probably automate that one day.

> you enabled an LLM to do destructive things and then complain when there is destruction.

There has been no destruction in it's little sandbox that I care about. It created a faulty command that got escaped by the shell, and oh shit, nothing of importance happened, because even if that command had been 'rm -rf /' I just would have had to return to the last backed up git that I had. That takes about 5 minutes I suppose. The amount of effort that LLM has saved me over the last two months is probably about a year's worth of full time coding, so I'm going to look past any accidental nuclear holocausts where it turns all the sand in its sandbox to glass.

Look. LLMs are just tools. Some tools are great, some tools are shit. This particular class of tool is great at all that boilerplate stuff that can get you 50 - 80 percent of the way to completing a task that you need to do today, if you already have enough skill to know how to do the task. It's been quite useful going through all my old projects in 'code review' mode, because it can often see security/usability/structural issues or improvements that I frankly could not be arsed to deal with when I wrote them. It is in no way ready to replace a human for any sustained length of time, and anyone with a shred of intellect can see that.

Personally, I don't think the AI bubble is going to last much longer. So I'm quite happily burning up Anthropic's compute time finishing off/cleaning up heaps of small home and work projects before a single question costs $8.25 for the LLM to answer.

the vehicles start looking more like a wingless airplane than a car.

And what does the company's name mean? ;) (Hint: it's Greek)

It adds up to 40 miles per day, not 5. 5 would be a cloudy day.

I used to be anti-solar-panels-on-cars back when solar panels were expensive, and ones of reasonable efficiency were even more expensive - the argument was, "put it on your roof where it belongs". But that's just not the case any more. Adding solar is just not that much of a cost to the car. It adds some complexities, but mainly in the design / early manufacturing phase.

Also:

The average American drives 37 miles per day.

1) So if you're in a sunny climate, it covers all of said average-driver's driving. Otherwise, said average-driver has to plug in occasionally, but not nearly as often.

2) Most people drive less than the average (the average is skewed by a long-tail - small numbers of people who drive very far every year). What you actually should be meaning is the median US driver; the median drives 23 miles per day. Most Europeans, even less.

3) Even for said "average american", their daily average is skewed by long drives (e.g. road trips and similar). Wheren of course you're plugging in, you'd be plugging in even if the car was adding 80 miles a day. But when not on road trips, their daily average is lower.

4) Surely you can see the appeal of the tangential benefits, such as being unstrandable - where even if you run out in the middle of the desert 20 miles from the nearest town, you're still going to get there, just delayed (remember that EV ranges, if you drive very slowly, increase like 2x, so 40 miles a day becomes 80, so a 20 mile shortfall is only a ~4h delay on a sunny day).

5) Nobody is saying, "One car for everybody". Of course appeal varies by person and by location. Here in Iceland for example we have three problems. One, very little sun at all for a good chunk of the year. Two, even in the summer, when the days are long, the sun is mainly low and circles around you. Solar power just kinda sucks here in general. And three, the three-wheel config would mean that the centre wheel wouldn't align with tracks in the snow from other cars (although there is a slight advantage, in that it also wouldn't align with road ruts from studded tyres, which often fill with water in the rain and become hazardous).

But somewhere in the southern US, it's a great option.

I'm seeing more and more references to "Honest" in AI output, or AI-related comments.

The AI didn't make an "Honest" mistake. It does not have the capacity for honesty. The output from a LLM is phrased in such a manner to provoke empathy, in a similar way to how Microsoft re-jigged all their user interaction dialogs to include "We" to soften the blow of their crappy software failing the user for the 5th time today. (Side note: "Something went wrong" is the most infuriating error message ever.)

When I ask a LLM for a code review it often blurts out "Honest note:" about some shortcomings. I don't care about "honesty". I care about safe, working, robust, code. The fact that LLMs are tripping over themselves trying to be "Honest" about mistakes in their "path of most statistics" output is a concern if you care about trying to make them operate outside their sandbox in the real world.

Yesterday Claude quoted a word in backticks during an automatic git commit and my shell escaped it tried to execute it. Luckily the word was just an English word with nothing matching in my path. But this is basic, basic, basic stuff. It's been committing things to git ever since it was built, and yet, it keeps tripping over itself. In my code one of the tests keeps failing due to seed data timestamps not lining up with the datetime the test was run. I can see that. Every time Claude runs the tests, it burns up tokens going, "Oh this particular test failed I'll just dig into things and see what's going on, **$$**$$**$$** oh it's just a timestamp issue". Never once does it commit that to its memory file, so eventually I told it to remove the test, and it just added a comment to it saying "Ignore this test due to timestamp misalignment", which it could have done the very first time, if it actually had a brain.

LLMs are a very handy tool if used right. I can get huge chunks of boilerplate code out of them with just a few sentences and that's great when I'm hashing out a concept. But to promise the world (and your investors) that LLMs are ready to replace people out in the real world, where "Honest Mistakes" have Real World Repercussions, that's outright fraud at this stage.

> It will randomize over a surprisingly short time if it doesn't have power for data maintenance.

If you plug in a USB stick, you just power up the controller and the flash memory chip. You have to read everything and write everything to charge the cells back up again, there is no background refresh going on like in DRAM.

For those who didn't follow it, it's not that it's a contact binary that is so neat in and of itself, it's that when they modeled it, they determined that, the collision that formed it was less than 5 meters per second (less than 11 mph / 18 kph). Like a parking lot fender bender, but with the cars being ~750 billion tonnes.

Yeah, there's two main problems:

1) People entering the wrong fields. For example, medicine really needs workers, at all levels, but not enough people are going into it.

2) Certain manual labour fields, like field work and home construction, because... well, I think we all know why there's a shortage of workers in those fields.

Or let's put this another way. Show of hands - how many of you "spicy autocorrect" / "stochastic parrot" people had "AI will start mass-solving Erdos problems" on your forecast list a couple years back? Huh, none of you? Fascinating!

Take some time to reassess your priors. And while you do so, understand that, yes, they are doing logic / reasoning.

They weren't discovered by an LLM. They were known conjectures that were proven by an automated solving language that was linked to an LLM.

I'll take "Things That Didn't Happen For $200", Alex.

Only a handful of meaningful proofs have ever been done by automated formal theorem solvers (the Four Colour Theorem being the most noteworthy example - but its proof is so long that humans can't verify it). By contrast, AI tools have been solving Erdos problems en masse. The majority of them just bog-standard commercial models. In case you need help, the only ones on that list that were hybrid (AI / non-AI) in the actual solving phase are:

1) AlphaProof / DeepMind Prover Agent / AlphaProof Nexus
2) Aristotle (Harmonic)
3) Seed Prover / Seed Prover 1.5 (ByteDance)
4) AxiomProver (Axiom Math)

In each of the above, LLMs come up with the lemmas / strategies but then use Monte Carlo search ("brute force") or likewise to investigate what they came up with. These are a minority. In the "AI Standalone" category, these "hybrid" tools made up only ~20% of attempts and successful proofs. Hybrid tools actually made more of a contribution in the "AI Alongside Literature" (related literature found afterward) and even more of the "AI Building On Literature" (related literature known beforehand) categories, which is the opposite of what people like you expect.

And even with the hybrid tools, it's still the AI doing the heavy lifting when it comes to strategy. Non-AI theorem solvers, again, don't have a spectacular record for churning out novel proofs to unsolved problems. Tools like Lean are more about mathematical rigour - a passive environment that requires a driver (a human or AI) to feed it actual strategies, lemmas, and proof steps. And no, you cannot brute force "strategy" in the vast majority of cases, which is, again, why automated theorem solvers don't have much of a track record with unsolved mathematical problems.

Let's take a random example: the disproof of the unit distance conjecture. It was solved purely by a general purpose commercial GPT model, not custom-trained to mathematics, with no external tools. Read what the various mathematicians reviewing / commenting on it have to say (sections #3 and onward). Seriously, don't skip reading them, actually read them. This was one of Erdos's favourite problems. He mentioned it commonly in his lectures. Essentially every mathematician working in complex geometry has thought about this problem. The approach that the model came up with was highly novel approach, based on CM-fields and class field towers.

I know you don't want to accept this reality, but it is the reality, so you better improve your ability to accept it,. The field of mathematics is already doing so.

Dang, link didn't post.

Ah, good 'ol "Model Collapse" theory that people have been pounding on for years now, predicting an imminent collapse in model capabilities. How has that been working out for you?