Tadpole9181

Until there is.

We have thus far been supremely lucky that irrational actors have not acquired a nuclear bomb. But the reality is that throughout history most leaders have been irrational.

My country (America) is behaving increasingly irrationally. I've been hearing fringe conservatives calling for nuking the middle east since the early 2000s, but the rot has spread and became deeply religious.

Now I hear talk of the end times from the most powerful people in the country, actively building bunkers. Groups of Republicans openly talking about the US being allowed to take whatever we want from anyone we want through war (Venezuela, Iran, Greenland, Canada).

Almost all of our leadership has been replaced in every branch of government. SCOTUS is clearly compromised and bought. The legislative may as well be dead. Almost all executive agencies have been purged, what remains have had their heads swapped with sycophants. Something like half of the military generals have been turned over. The DoD was renamed to the DoW with cheers, as an alcoholic prison guard was put in charge and has started touting hypermacho religious warrior catchphrases and openly advocating for war crimes.

The other day we testimony from judges before legislature who refused to say if Presidents can run for three terms or that Donald Trump lost the 2020 election. Under oath, just simply refused.

And that's before getting into how the President of the United States started a war to distract from his involvement in the highest profile sex trafficking ring in human history, and has continued to use his position to perform insider trading on oil prices repeatedly for weeks.

If you said any of this would be happening in 2005, I would have laughed at how "that would never happen in the US". And now it's just... reality. Like Russia. And it seems to have no end in sight. So what actually happens if Trump gets so deranged he hits the red button? And if not him, who will the loonies pick tomorrow - someone who actually believes in "the cause", who doesn't understand it's a grift?

And what are the external long term consequences of Trump's actions in countries like Iran and Venezuela and Ukraine? The US is not to be trusted anymore, every country has a clear signal to build their own nuclear arsenal ASAP. We have a river of blood behind us, so are we just expecting to get lucky that no one in Iran, for instance, who gets to a position of power leverages an irrational hate for the US and just... Does it in a religious or hate-driven Armageddon?

From your own link, there are 12,000 injury related deaths. Of those, over 1/3 are firearms related and 1/3 of those are homicides. They may not be school shootings, but damn does that seem related to the second word.

Anyway, to answer your question: because we already do a lot to stop the other causes. Cars have oodles of safety regulations, for instance. Home pools have rules on covers in many states, lifeguards are placed at water features. But we can't just snap our fingers and stop all car deaths, or cure all disease.

Meanwhile shooting deaths are an entirely artificial problem, which we actively do less than nothing to solve. Both by avoiding root cause fixes - not only avoiding addressing mental health, but actively encouraging bullying of minorities like trans students - but also avoiding regulation of the most effective tools (guns).

And, just a final point, humans are human. A kid dying in a tragic accident or disease is heartbreaking. But a dozen kids being eviscerated on Live TV repeatedly for someone's entire adult life, with tens of millions of people having the audacity to say "get over it, we refuse to talk about this" or "even "it's a hoax"? That's enraging.

We are emotional animals, but I'm more than happy that we try to leverage our empathy to make the world a better place. We can do more than one thing at a time, after all. For instance, under your argument, nobody should have bothered inventing amber alerts because it doesn't save hundreds of kids per year.

Beyond the obvious fact nobody really uses the cold war definition anymore, are you really insinuating those countries are socialist?

This logic never made sense to me.

The woman is clearly a human life too, and every western country has a right to self defense. A fetus is an imminent danger to women, their health, and their property. They have a right to remove it from their own body to stop that.

Or am I allowed to unconsentually put myself inside anyone for months, eating the food from their blood like a tapeworm, causing chronic pain, permanently changing their hormones, potentially risking death, just to come ripping my way out of their body and then demanding 18 years of pay and attention?

Am I entitled to an unconsenting healthy person's blood or organs or marrow if I'm going to die without it and we know they're "probably" going to be fine?

If multiple people are confused on what you're saying, you should probably clarify with better wording.

So you don't actually have real criticisms of the architecture at all...?

I mean, insult me all you want, but you'd be pissed if your mechanic didn't use code readers or your carpenter used hand drills to pad their hours.

I do hand programming for myself now and use the right tools in my office, because I'm a responsible adult who understands that I am not paid to have fun or feel special and smart - it's to produce a product for my employer.

Why? I'm an "artisanal programmer" at heart, but for the past year I've been using these to great effect. A car that needs to be started with a hand-crank and occasionally stalls is still a car.

Let's be clear, there is no underlying crime at all as far as we are aware. So the appropriate hypothetical is:

You ate a grilled cheese in your home and wipe the crumbs into the bin. The police arrest you for destruction of evidence for robbing a bank that never even got robbed. Then they hit you with qualified immunity and high five each other.

It's utterly ridiculous that police seem to just have infinite powers now. All they have to do is say "I suspect" and they have boundless ability to arrest anyone (potentially losing their job or missing key life events), declare any object as evidence (which the department can conveniently keep for themselves even if you're found innocent), or use mass surveillance systems (to stalk women).

EDIT: Finding another article, it's even worse. He didn't do it at all - he was interrogated, was never allowed his lawyer, never read rights, with no warrant. And after having the device seized an agent triggered the device wipe themselves by trying to get in under this duress.

So in this case, the the cop eats the grilled cheese and throws out the crumbs and then arrests you for destruction of evidence.

ICE has openly murdered multiple US citizens in broad daylight and on camera, hundreds of miles away from any border, with zero consequences.

You're starting to perceive us as totalitarian? We are! For 30% of us, proudly, apparently. Our president attempted to violently overthrow our government for God's sake!

Something like 90% of US residential buildings have AC. For Europe, it's around 20% (for good reason, mind you). It's not a meme if it's an actual fact that many European homes are not prepared for these kinds of increasingly common heat waves, and I got the impression that the parent was showing genuine concern.

I disagree, this is all because it's understood.

You need a system promot because an LLM is fundamentally a token predictor. It needs to be primed for the work it's going to do, otherwise it's next token prediction has too little to go off in the beginning and goes nuts.

You use a separate AI to review the code, because it's a token predictor trained mostly on accomplishing tasks in a cost effective way - we haven't made AGI here. The first shot that actually makes the code will have rationalizations in its memory as well as prior research, which bias token prediction to accepting that as true (remember how they had to train our sycophancy?). Then because of the desire to be cost effective, they're slightly lazy, and so it won't always do the research to find new edge cases and problems and missing tests the initial research didn't find.

So spin up a separate, clean slate, and ask for it to review from scratch. Or have that one spin up multiple smaller ones to have them specialize in specific concerns or domains (security, data model corruption, code quality), then have the orchestrator validate those concerns and stitch them into a cohesive response.

I can't help you on your last question. Saying they are degrading in quality is not even close to my experience. 5.6 and Opus/Fable 5 are have been a huge step up. Though I do need to adjust memory rules as these new models come out, since ground-up retrained models often come with their own quirks that replace old ones - causing old, specific memories to have unintended side effects.

First of all, this is actually how this part of the conversation started:

> If the US companies need trillions to barely beat Chinese companies spending billions, despite a multi year head start...

Because they're talking about the cost difference of distilled model development and ground-up trained model development.

And second, the answer is that OpenAI and Anthropic had to do all the research into how to train models. Then they had to acquire all the data, curate and filter it. Then they had to design all the ways to iterate on training and antagonize it to be better - because there's not actually an enormous corpus of aligned, human stream-of-thought data. Then over half a decade they've been refining these methods.

There's no way you don't understand that if it was not easier and cheaper to distill a model, the institutions in question would be training their own models from scratch.

Can you please stop being coy and intentionally obtuse? Just have a discussion in good faith, I'm so beyond sick of this kind of rhetoric.

Yes, AI training uses human data and a lot of it was not compensated. But that has absolutely nothing to do with the thing this thread is about.

Distilling models costs less money than a really procuring quality data and training a model yourself. If you disagree, debate that.

After going to the official website, I actually spent a couple of minutes looking around and checking certificates to see if it was a scam. The page makes it sound like it was built by a single guy who isn't incorporated, but if you go to the Terms, it says Sophon LLC - who seems to do AI document processing? But then going to their About page and "Our Team" is some cagey "experts from leading companies".

Combined with the Apple button reading "Download for Windows" on a Windows machine, and basic HTML issues like `we're`, I'm not inclined to have this software running my entire PC.

To be clear, did or did not the party leader say he would like to deport people born in Sweden to legal immigrant parents because they are not "naturally Swedish"?

> I'd remind you the current new version is not an improvement compared to the previous one, both in terms of correctness and maintainability.

Except, you know, multiple real companies saying that it is and using it in production. And the fact it closed all known memory leaks. And that nobody has a really pointed to a single actual issue the new version introduced after two months of endless, ceaseless bitching.

I'm also fascinated by all these people upset at Jarred for harming the readability of a codebase they've literally never cared about before it become drama. Bun has almost exclusively been maintained by Oven employees since it's inception.

> Deno for instance have ~0.2x as many unsafes.

A project written ground-up in Rust idiomatically, with a smaller surface area, still has an unsafe footprint within an order of magnitude compared to this automated rewrite from an unsafe language with different practices and known bugs? That's not exactly the slam you think it is.

It isn't disabled, it has exclusions. Now that they have it, they can close the gaps. It was literally impossible to have ANY coverage before, now they are mostly, covered and have an avenue for remediation.

I don't understand why folk are having such a hard time understanding why you do large projects in multiple steps? 80/20 rule? Perfect is the enemy of good?

Was nobody here for moving billions of lines of JavaScript to Typescript? It starts with declarations, then turn on type checking gradually inside the codebase: piece by piece until done.

Just so everyone know, Bun is and has always been owned by a YC funded org with full time employees. The overwhelming majority of commits came from them.

This feels like such an absurd, bad faith take I keep hearing.

In Zig, every single memory operation is unsafe.

And Bun must interface with C code that has no safe interface, necessitating a ton of boundary-level unsafe behaviors.

There's too much, sure, but can we at least be honest and reasonable?

The canary build has been Rust for over a month, available to anyone. In that time it has been used in production for Claude Code and Prisma Compute.

> and from my own experience it's very frustrating to leave a Claude session running and come back to find it did nothing because it got stuck on a question.

I cannot fathom implementing and shipping a feature to hundreds of thousands of people without even asking basic questions like: "what types of questions does Claude ask users".

Literally one of the most used plugins in their entire ecosystem, provided via their official plugin marketplace, is Superpowers. A plugin whose very first operating step is _asking numerous questions about product requirements_. Of course those prompts can't be skipped!

It wasn't even parameterized for Claude to tell the prompt what severity of question was being asked to allow at least _something_ to categorize urgency or expected response time.

Even more egregiously, 60 seconds!? The first time I noticed this happened was when it asked me a question, I turned to my second monitor to go look at some product documentation to get an answer, and by the time I turned back it had skipped me. How can I possibly provide any kind of informed answer in under 60 seconds? I can barely read some of its context for a question in 60 seconds!

I don't think they did this with malintent, but I do think this shows an enormous gap in judgement in how they handle the idea to delivery pipeline.

Trump fired all members of the Election Assistance Commission this week. His head of the USPS confirmed they have a proposed rule to not deliver mail-in ballots in select states. Last night, as we approach midterms, Trump gave a speech saying US elections are illegitimate and only he can fix it (by controlling the voting machines and who gets to vote).

I'm really sorry, but it really does look like we may have had our last actual election already.

FWIW: "Baba Is You" is 7 years old and heralded as one of the greatest puzzle games of all times, with guides and solutions shared all over the internet. How to beat this game is 100% in the training set.

To quote the people who make it:

> ARC-AGI-3 is an interactive reasoning benchmark which challenges AI agents to explore novel environments, acquire goals on the fly, build adaptable world models, and learn continuously.

This harness does nothing to actually accomplish those goals.

It's a clever trick, sure, but you aren't allowed to use a calculator on your basic algebra tests in school for a reason.

This is not actually running the Arc-AGI-3 anymore. To summarize TFA:

1. The AI plays the game and records outputs.

2. The AI does TDD using those outputs to create its own copy of the game.

3. The AI then uses it's copy of the source code to understand the rules. This bypasses the intent for Arc-AGI-3 to test the underlying model's ability to intuit game rules naturally, like a human.

4. The AI then runs simulated moves on the copy of the game before playing them in the "real" game. This bypasses the intent for Arc-AGI-3 to test the underlying model's ability to plan and predict moves, and track world state in its "head" over time.

To make an apt comparison... You go to get your chess ELO. You don't know chess at all and you're really bad at it, so you pull out your laptop and write a chess engine. Then when you go to get ranked, you just copy the moves from the software. Now you're a grand master.

The point of Arc-AGI-3 is to measure model performance. We already know that models can one-shot and iterate on very rudimentary game implementations. And, naturally, once it effectively has a copy of the source code, it can use that to play the game better.

This harness is really moving the goalpost by defeating the entire point of the test. Instead of seeing the strength of a model's world view, its ability to internally derive and intuit rules, and its ability to keep track of game state over time, we're just letting the AI cheat. This is just the LLM equivalent of running a chess engine to the side.

And this harness would not work in a remotely complex game and relies on the fact that Arc-AGI-3 is a focused test that only made the games as complicated as they needed to be for current model performance.

> If it's something that anyone could do, without skill, why would anyone be impressed by that?

Gifts aren't supposed to be impressive. Nor is their value measured by effort or cost.

Nor does this reply answer their question of "is hiring a professional to to it any better?" It seems you would have to argue no, though.

What exactly is your complaint with it?

`auto&&` has been standard syntax in C++ for going on 15 years now and has a very clear, irreplaceable meaning - a type-deduced forward reference - to any remotely competent developer.

Except the blog post shows that they fixed a hundred or so known issues, patching several memory leaks and making the project viable for Prisma Compute's adoption - which it wasn't before. It's now running in production in two places just fine.

Can you point to an equal number of issue tracker tickets showing novel bugs or regressions in the canary build?