thorum

IANAL but I read that as forbidding you to provision legal/medical advice (to others) rather than forbidding you to ask the AI to provision legal/medical advice (to you).

createElement(‘tr’) and table.appendChild(row)

Actual link seems to be: https://genai-showdown.specr.net/image-editing

The source code is not the LLM. The LLM is billioms of random floating point numbers that somehow encode everything the model knows and can do.

The ML field has a good understanding of the algorithms that produce these floating point numbers and lots of techniques that seem to produce “better” numbers in experiments. However, there is little to no understanding of what the numbers represent or how they do the things they do.

Am I the only one who has simply said “no thanks” to Windows 11?

There were hints of where Microsoft was heading in Windows 10, but at least a lot of the worst “features” could be disabled.

I find 11 just completely unacceptable software to run on any system I own.

It’s long been theorized (not proven) that fully balancing the microbiome in IBD patients might interrupt the disease cycle and lead to remission.

"not a X, but a Y" - 8 matches

"is more than a X... it is Y" - 3 matches

"not just X, but Y" - 4 matches

The article’s advice section is actually quite practical, and directly addresses your point:

> It should go without saying that the best way to increase your practice surface area in a given field is to be obsessed with that field. Obsession makes quick work of formal and bounded training sessions, and it doesn't need "tips" on how to do so.

> So the question then becomes, "How do I increase my practice surface area if I'm not already obsessed?"

I disagree, so not everyone, I guess!

I would argue that we just get pickier and more sensitive to slop. When everyone can make a movie, the standard for a good movie will be higher. Many current Hollywood films wouldn’t make the cut. But maybe some kid in Nigeria makes the greatest film of all time.

People are doing cool things with it. Here’s one example:

https://www.tiktok.com/@dreamrelicc

Before AI, each video on this channel would have taken a large team with a Hollywood budget to create. In a few more years, one person may be able to turn their creative vision into a full-length movie.

I also had unsupervised internet as a kid in the 2000s, but the internet is a very different place now with social media, algorithmic feeds, and content that has been carefully designed by bad actors to be as addictive and manipulative as possible. AI is surely not helping so far. I think it must be very difficult to grow up immersed in that environment.

If in a few more years, “video games” evolve into hyper realistic VR worlds with LLM-powered NPCs that act like real people, also designed to maximize engagement, we might need to revisit the risks of exposing kids to that as well.

As a developer, this is how I feel when I open a higher math textbook. Even though I understand that, in principle, mathematicians are using a syntax and lingo that shares some common features with computer code, my eyes slide over the words and symbols like they’re a foreign language.

Sure but that’s exactly the “lost knowledge from a long gone advanced civilization” situation that this article suggests doesn’t happen.

> So: loss of technology is not impossible. But to an innovative and large culture like modern human civilization, it’s not really something that happens. It’s just a fun trope for stories. Let’s hope it remains that way.

I agree with “let’s hope it remains that way.” But you may be underestimating the fragility of modern human civilization and all the systems that are so effective at preserving knowledge. At the scale of history, most of our technology has only existed for the blink of an eye. It would not take much, really, for all this to go away.

Absent the internet, power grid, and other systems of global infrastructure, absent anyone to take care of the libraries (which are falling out of favor), absent an authority to gather together all the experts with their tightly focused specializations, absent the chip fabs and the global supply network, how exactly do you build a computer? Or even something as “simple” as a pencil?

A coding agent that receives no updates will be useless very soon. In a few months they’ll tell you to switch to Codex CLI, and that will in fact be a good recommendation.

Is the NYT crossword so much better than alternatives that it would be worth paying for?

> Six of the eleven picked the same movie

This is surely the greatest weakness of current LLMs for any task needing a spark of creativity.

My point was more “humans are used to tools that don’t always work and can be used in creative ways” than “no human invention has ever been rigid and reliable”.

People on HN regularly claim that LLMs are useless if they aren’t 100% accurate all the time. I don’t think this is true. We work around that kind of thing every day.

With your examples:

- Before computers, fraud and human error was common in the banking system. We designed a system that was resilient against this and mostly worked, most of the time, well enough for most purposes even though it was built on an imperfect foundation.

- Highly precise clocks are a recent invention. For regular people 200 years ago, one person’s clock would often be 5-10 minutes off from someone else’s. People managed to get things done anyway.

I’ll grant you that Roman aqueducts, seasons and the sun are much more reliable than computers (as are all the laws of nature).

I like this framing, but I don’t think it’s entirely new to LLMs. Humans have been building flexible, multi-purpose tools and using them for things the original inventor or manufacturer didn’t think of since before the invention of the wheel. It’s in our DNA. Our brains have been shaped by a world where that is normal.

The rigidness and near-perfect reliability of computer software is the unusual thing in human history, an outlier we’ve gotten used to.

I don’t know enough about statistics to answer these with math, but I’ve been on quite a few buses and it’s common at some stops for bus arrivals to cluster around specific times. If you always leave after first you see, and most of your random observations are before the first bus, won’t you (almost) always miss the others?

I’ve been trying to convert a dense 60 page paper document to Markdown today from photos taken on my iPhone. I know this is probably not the best way to do it but it’s still been surprising to find that even the latest cloud models are struggling to process many of the pages. Lots of hallucination and “I can’t see the text” (when the photo is perfectly clear). Lots of retrying different models, switching between LLMs and old fashioned OCR, reading and correcting mistakes myself. It’s still faster than doing the whole transcription manually but I thought the tech was further along.

This article is insightful, but I blinked when I saw the headline “Reducing the human bottleneck” used without any apparent irony.

At some point we should probably take a step back and ask “Why do we want to solve this problem?” Is a world where AI systems are highly intelligent tools, but humans are needed to manage the high level complexity of the real world… supposed to be a disappointing outcome?

I think we’re going to have to set bigger goals for ourselves. We’re all still figuring out what that means.

We need a new set of UX principles for AI apps. If users need to access an AI feature multiple times a day it should be a button.

> npm scripts make a http call to the npm registry each time they run, which means they are a terrible way to execute a health check.

Is this true? I couldn’t find another source discussing it. That would be insane behavior for a package manager.

The Information’s report from earlier this month claimed that GPT-5 was only developed in the last 1-2 months, after some sort of breakthrough in training methodology.

> As recently as June, the technical problems meant none of OpenAI’s models under development seemed good enough to be labeled GPT-5, according to a person who has worked on it.

But it could be that this refers to post-training and the base model was developed earlier.

https://www.theinformation.com/articles/inside-openais-rocky...

https://archive.ph/d72B4

The quiet revolution is happening in tool use and multimodal capabilities. Moderate incremental improvements on general intelligence, but dramatic improvements on multi-step tool use and ability to interact with the world (vs 1 year ago), will eventually feed back into general intelligence.

The product simply should not be called Autopilot. Anyone with any common sense could predict that many people will (quite reasonably) assume that a feature called Autopilot functions as a true autopilot, and that misunderstanding will lead to fatalities.

If you’re a power user of these LLMs and have coding experience, I actually recommend just whipping together your own bespoke chat UI that you can customize however you like. Grab any OpenAI compatible endpoint for inference and a frontend component framework (many of which have added standard Chat components) - the rest is almost trivial. I threw one together in a week with Gemini’s assistance and now I use it every day. Is it production ready? Hell no but it works exactly how I want it to and whenever I find myself saying “I wish it could do XYZ…” I just add it.