daveguy

Try Kagi. It's definitely worth it. And you can block sites in search very easily. None of the closed content farms show up in my search results.

They must have edited out the 1 in 5 times the agents crap the bed and mess up something important.

Our government is an embarrassment and a disgrace right now. Anyone who wasn't guzzling elon's twitter or fox's hatefest knew it was going to be this way. We need severe regulations on oligarchs and their propaganda toys. Used to be we had enforced anti-monopoly controls on media outlet conglomerates. The anti-government Goldwater/Reagan legacy worked for decades to destroy regulation. It needs to return with a vengeance.

IIRC there are already FLAs. FLAs too. ROTFL.

Well, we could add a + after any acronyms after they have a formal definition in TLA+. Except for TLA itself. That would need to be TLA++.

I'm curious what function a data diode (unidirectional network) would provide besides sensor data/mirroring/replication. Once you do anything that requires packet confirmation (TCP) you open yourself to OSI layer 4 security risks. Seems like mirroring/replication would need some feedback and then is it just sneakernet restore? Or another diode out from the replication for specific processing? If the same control system has access to both diodes then it's not a unidirectional system anymore. Is a data diode more of a pseudo-unidirection where it is enforced above TCP?

I completely agree that IT admin could be a lot more secure by design. Combined with better interfaces for responsible configuration.

Clearly we need to update the CFAA to include criminal prosecution for negligence.

Pretty sure that was the argument de jour when OpenClaw came out.

Thank you!! I saw this video about a decade ago, couldn't remember where it came from, and was having trouble finding it again! Bret Victor is an HCI juggernaut. From this info, here are some other links:

his website: https://worrydream.com/

his page for learnable programming: https://worrydream.com/LearnableProgramming/

There was a programming environment closely related to his work that was a kind of visual database. But it may not have been developed by him / his lab. I don't see any obvious links to it on his website.

But you should see how great it is at crud social media apps and 2D scrolling games!

I guess that depends on how many robots a company is willing to have stuck in random places.

That's a great point. A training corpus based on written text will be inherently biased toward confident and right. Then the RLHF exacerbates the problem because people respond more positively to confident and too often assume correct when they read a confident response.

Oh they don't just fear the optics. They actively work to surpress negative stories on hacker news.

> This makes it sound like RL rewards a confident tone -- in general, I don't think this is true (most RL is RLVR, which typically uses binary verification of correctness).

A binary response vs rating is not related whether it learns confident or hedged tone. Either will produce a confident tone because humans respond more positively to a confident tone, hence the conman's language. Binary or not humans reward the tone and very much bias the model.

But there's an even more contrived reason the training set contributes. The vast majority of human writing is confident. When the prior is greatly biased, a random number generator biased to that prior does better. The difference with humans and machines is humans are less likely to respond if they are less confident because they understand not knowing, which is why the training set is biased. It is one of the many fundamental flaw of LLM training and confusion of LLMs with intelligence. And that will not be fixed within the LLM architecture.

> But I’m not going to fucking babysit a machine, sorry.

A machine that does not learn from the feedback. The only thing these LLMs know about after the initial training is what makes it into the context. Imagine if someone couldn't learn anything new without someone else explicitly telling it what was important to the task every single time? You wouldn't call them a junior engineer.

So is a random number generator. That doesn't mean it can play Pokemon.

> A pre-trained universal classifier that can replace specifically-trained ones would have been considered just as much science fiction in the 2010's as the capabilities of modern LLMs. I'm not sure Jev is actually there yet, but at least it sounds theoretically doable today.

That's the point. It can't. And it's not even close.

Doesn't matter if it was a partial stolen solution instead of a complete stolen solution. It's the stealing part that matters.

I'm going to guess part of the influence campaign is getting stories like this removed from the front page of HN as quickly as possible:

https://www.technologyreview.com/2026/09/22/1144867/dont-be-...

eg:

https://news.ycombinator.com/item?id=49814211

Not even flagged, just disappeared.

Well, this one got disappeared from the front page fast.

Spoiler: openai could neither confirm nor deny that the model that "solved" navier stokes was trained on the chat logs of the mathematicians who did solve it using a similar technique.

1) Those problems did not go away. They only "went away" if you assume no one will ever need to read code again. And if you believe that, you clearly aren't reading the code already. Which I guess is why you think naming and architecture problems went away. AI is the absolute worst at architecture decisions.

2) not even sure what you mean by this

3) probably shouldn't admit that. It implies the reviewer has a less than average understanding of the code they're reviewing.

4) preventing and managing vibe code devolving into a pile of slop requires programmers not use AI. 80% accuracy repeated in more and more layers === more and more failures. In other words, the skill required is exactly the skill of being a good programmer without AI.

5) e) no AI all the time or only use AI as search. You're almost there with a and b. With c, it just doesn't understand well enough to "quiz you". With d, how are you going to know if the docs are correct if you aren't reading the slop?

Hahaha. I remember when musk and his merry band of sycophants were bragging about grok producing the only physically accurate bicycle. What happened?

Maybe a better approach would be to stop trying to get a product out of a slot machine?

LLM chatbots are software designed to manipulate, addict, and mine data. Just like social media before it. But anyone who bothers to read the output in a domain they understand will discover they aren't all that no matter who OpenAI steals research from.

Please tell me they are going to have independent computer scientists and security experts to ask questions that expose the doomerism/"rationalist" LessWrong bullshit. Please don't just let him spew talking points designed to make a deflating bubble look like a good thing.

You'd think that believing they are going to destroy humanity would motivate them.

Seems these things about genAI are fairly well established at this point. Hence the vast majority disapproval of genAI being forced on users and broad rejection of AI data centers. Did you have a substantive response for any of the three major negatives, or just your own low-effort, low-care dismissal?

It's all of the above. The unreliable empty garbage they barf out, threats to destroy jobs for oligarch asshat profit, and the resource hungry pillaging of neighborhoods. Oh, yeah, and also the crappy mush up of stolen art.

We have been able to stage reckless experiments with janky tools the entire time. Used to be it got you thrown in jail or at least a visit by the FBI. Says more about the tech bro dipshits than the tech, if you ask me.