Zambyte

LLMs are chaotic pure functions. Their input is usually randomized.

It's at least important from philosophical perspective regarding whether LLMs have free will or not. It's hard to say whether or not a human thought or action is driven by free will, people have debated it for centuries. We can provide the same input to an LLM and get the same output, because they are deterministic. Surely it is obvious that a pure function does not have free will, no matter how expensive it is to compute.

Traditional non-chaotic systems*

LLMs are deterministic. They are chaotic, which people confuse for non-deterministic.

No problem. I think it's less of people withholding information to have an advantage, and more people still not having settled on a workflow they like. Herdr is the most recent addition in my workflow as of only a few days ago, but it directly solves problems I have been having (juggling tons of terminals, even with my tiling wm has been a little unwieldy). The rest I've pretty much settled into for a while now.

Up until the last couple of months, I have treated LLMs as a supercharged stackoverflow. I would ask it questions on how to do something in a general sense, and then adapt the answer to my use case.

Now, my entire programming flow does not even include an editor. The tools I use are: pi.dev to write and implement openspec specifications, herdr to manage many pi instances, and ollama to run qwen 3.8 27b on my single 7900 XTX.

Writing good specifications is the key detail here. I will often iterate on a spec for hours until I am happy with it all of the details. Once I am happy with the spec, I can be quite confident that when I tell pi to apply the spec, the changes that I want will be done, and done how I want them, when I come back to check when it reports itself as done.

The landscale is fundamentally different from what it was. Feel free to ignore it, but you can absolutely generate high quality code if you know what you're doing.

Well that's what they're doing.

robots.txt is a suggestion, not a rule. It's a service to crawlers to help them avoid wasting time. Using robots.txt as a security measure is like trying to stop a foreign invader with a "road closed" sign.

Edit: this is obviously assuming they literally meant robots.txt, but from a qyick skim of the site, it doesn't look like that was mentioned at least. I guess you meant it metaphorically :P

I like that you specified that the car was self driving, but didn't specify the RV was.

When sitting, the optimally comfortable angle to look, and the optimal place to hold a controller are very far from each other. The Deck is basically only fun for long sessions if you can play it laying down. VR is more flexible in this regard.

The whole point of butchered websites is to direct people to install their more invasive app. It's not like all companies just got bad at making websites somehow. There are conflicting incentives.

To people downvoting: How do you define "general intelligence" in a way that AI currently is not able to do in some way, but humans are?

AGI isn't coming, because we've had it for years. People just expect AGI to look like sci-fi, and hold the artificials to higher standards of "general" and "intelligence" than the naturals.

Luckily, AGI is pretty boring. It's a tool that does it's job. It seems like very wishful thinking to expect the same of ASI.

They did more than let them. In an abstract way, they told them to. They gave it all of the training data it had at that point, and then it did the thing it was trained on. Of course they should be help liable for programming their computer to hack another company without permission. It doesn't matter that they spent a lot of money doing it.

I haven't noticed.

The advantage of busses and bikes are that they can use the same roads we already built for cars. We don't need to build new infrastructure to make busses and bikes much better, we just need to limit car access. That can be done by significantly lowering speed limits for private vehicles, tolling a lot more roads and streets, and in cases where we can (especially in cities) ban personal vehicles entirely.

Trains need infrastructure, but a lot of the train infrastructure is already there, it just needs to be repaired and modernized. The USA freight network is one of the most comprehensive train networks in the world. The problem is that is has gone through decades of neglect, and trains are required to crawl through a large portion of the network so as to not derail. Urgent modernization of the rail network could happen in years instead of decades.

Yup, same here. Been using Kagi for years. It's boring, it just works. Hoping they can stay that way.

I'm American, and I don't normally drive, because I use busses, bikes, and trains to get to my destinations. I sold my last (hopefully ever) car over a year ago.

I'm sorry to shatter the dreams of most Americans, but the world has finite resources. Buying a car when you're poor is hard, because cars are expensive, because they are extremely resource demanding. Even with all the subsidies drivers receive, they still usually struggle with the cost.

By the way, you don't have to look at public transit schedules when your public transit works. I recommend visiting countries outside the US that have functional infrastructure.

The solution to car problems is not more cars. It's serious alternatives to driving. We don't need Chinese cars. We need trains, busses, and bikes.

What about selling busses and trains, and buildings bike lanes and train tracks? Loans and cars aren't the solution to loans and cars, but there are solutions.

I like to think of it as a how -> what -> why pipeline.

Early programming in assembly mostly involved telling the computer how to do things that you want it to do. You instruct it in terms of the mechanical capabilities of the machine.

Compilers and interpreters came along and raised the layer of abstraction that programmers work on from the "how" domain to the "what" domain. Instead of saying how to move bits around to solve problems, we tell the computer what to do, and it would do it. Even languages like C in which programs seem like a description of how to compute something, they are more of a vague suggestion to the compiler, which will do what you told it to do, how it wants to do it.

Now in the era of agentic programming, the "what" isn't even the domain that programmers are operating in, it's "why". Why should we use these technologies? Why should certain architectural decisions be made to support future development? Why should protocols have or not have certain information in them? The relevant "what" questions are handled by the LLM, just like the "how" is handled by the compiler.

I really like using spec-driven development to capture the "why" as artifacts. I personally use the tool openspec, and I like it because it explicitly captures my reasoning in the documents it produces, but I'm sure there are other solutions. I also have a prompt that will generate a commit message for a change that captures "why" instead of "what", based on the conversation history. This is better than just describing what changed, because the diff already does that, and if you're looking at a commit to understand it, generally you'll be confused by the "why".

No matter how you do it, find some way to capture "why" as a material artifact. The "what" is the code that basically anyone can make now. Why will maintain your purpose.

Can you back that up with numbers or do you just expect people to take your word that cyclists have killed about as many people as WWII?

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

https://www.academia.edu/115164883/Car_harm_A_global_review_...

People would absolutely choose trains over flying if we actually had modern train infrastructure. NY to LA would be about 11 hours at 250 MPH. That's 100 MPH slower than the fastest passenger train, but seems like a reasonable estimate for an average speed including stops. The difference between chilling on public transit for 8 hours and chilling on public transit for 11 hours is not that big of a deal for most people. Driving is essentially not an option for most people for this trip too.

Now do China.

My apartment overlooks a three way intersection. Drivers block the box almost every light cycle in rush hour. The most common traffic offense I see bikers make is riding on the sidewalk, which they obviously do because they fear being killed by drivers.

And ideally enough mass to violently crush their neighbors.

Distilling is a massive achievement. I can run Qwen. I can't run GPT (TM). It's not a matter of X is better than Y. It's a matter of Y exists, X does not.

I wonder if the prevalence of adblock has helped make a lot of things like this possible. Like, it's hard to imagine people putting up with this without literally protesting in the streets. But the people who can't tolerate it have mostly gone through the effort of opting out, and those who can just put up with it. And thus, the frog boils.

Advocating vs blaming. Blaming people for using inefficient vehicles when you eat red meat (assuming you eat enough red meat to have more of an impact than the fuel) is hypocritical. Advocating for better alternatives to inefficient vehicles while eating red meat is not hypocritical.

Being flawed isn't the issue. Everyone is flawed. The issue is blaming people for problems like environmental damage, while at the same time needlessly causing more damage yourself. The problem is being hypocritical. Some people are certainly more hypocritical than others.

The vast majority of people simply don't need SOTA models. I run most of my programming on a single 7900 XTX with Qwen 3.8 27b. I'm more interested in scaling out to run more instances of the model in parallel, than I am in scaling up to better models.