actuallyalys

One thing that appears to have been lost between GPT-4 and GPT-5 is that it no longer reminds the user that it's an AI and not a human, let alone a human expert. Maybe those genuinely annoyed people, but it seems like they were potentially useful measure to prevent users from being overly credulous

GPT-5 also goes out of its way to suggest new prompts. This seems potentially useful, although potentially dangerous if people are putting too much trust in them.

It’s definitely not a full solution, but it seems like it would solve cases where having the full history of the large files available, just not on everyone’s machine, is the desired behavior.

Honestly these results may say as much about the classifier as they do about the data they’re classifying.

In this case, it appears to be just that the author didn't properly qualify the sentence (see their reply to my comment).

But certainly it does seem like a potential risk of LLMs that they will synthesize people's notes and early marketing materials into "experience reports" that describe benefits and uses that never actually happened. Potentially these could then become part of training data or future contexts via search for other people writing, further exacerbating the effect. We already see similar phenomenon with human errors getting magnified via word-of-mouth.

Okay, gotcha. Thanks for clearing that up!

> At this point, most tech writing shops are serving llms.txt files and LLM-optimized Markdown

I find this hard to believe. I‘m not sure I’ve ever seen llms.txt in the wild and in general I don’t think most tech writing shops are that much on the cutting edge.

I have seen more companies add options to search their docs via some sort of AI, but I doubt that’s a majority yet.

I basically agree with the author that the iPhone's camera is inferior to dedicated cameras, at least in the hands of a photographer who's learned to use them. To me it's striking that there's even a question. My first camera as a child was a cheap film camera[0], and the first digital cameras I saw firsthand, as cool as they were at the time, had even worse quality. Now smartphones have much better quality for people pointing and shooting and do it while crammed into a device that does many other things.

Now, I'd hate for dedicated cameras to go away. I love shooting on SLRs, digital and film. I see smartphone cameras not as pretenders to the throne but as democratizing tools lowering the barrier for entry and a great way to get shots when you don't have your dedicated camera.

[0]: for the record, the issue with the camera was that it was cheap and I didn't know what I was doing, not that it was film.

It's too late to edit my comment, but to clarify, I meant skepticism of these comments by the attorney general, not of the restaurant that hosted the show.

Many conservatives who target drag shows believe drag and cross dressing are inherently sexual or inappropriate for children so I think skepticism is warranted.

The TCPalm spoke to some attendees [0] who said the bar does ensure minors are not present for the adult-only shows, which take place later in the evening.

[0]: https://archive.ph/g4nIq

That explains why they're expensive, but not why they're more expensive than they were before.

Given the age group (mean age of 71), I wonder if it's partly an effect of people with mild cognitive impairment, which the study explicitly mentions are included, being more sedentary (i.e., the effect is backwards or bidirectional). You could also imaging other risks of aging increasing both sedentary time and harming the brain (e.g., social isolation).

I feel like it takes a human a month to write a novel or start up a company only of you’re talking about a very constrained version of the task. Like people write novels in a month — that’s the whole premise of National Novel Writing Month or NaNoWriMo — but they aren’t finished products, they’re first drafts.

Similarly, while I’m sure you could make good progress on starting a business in a month, it seems like that would take longer to genuinely complete from start to finish. Also, it seems like it’s necessarily a task that relies on external factors: Waiting for approval to come from various agencies, hiring employees, waiting for other parties to sign contracts, etc.

I’m personally doubtful that venture capital is a socially beneficial or fair process [0], but I’m inclined to agree with the author of the post that the “experiment” isn’t a good basis for that conversation.

[0]: Yes, I realize the irony of posting that here.

I sympathize with this mistake because people are often very tribalistic, at least in certain online spaces, so it can feel useless to hear people out, especially when you don’t feel they will reciprocate.

That doesn’t justify it, of course, but I think it partially explains it.

Crime is also down compared to where it was if you ask people directly [0].

[0]: https://ncvs.bjs.ojp.gov/multi-year-trends/crimeType

Gotcha, thanks for clarifying!

To be clear, I'm only saying the unit is correct, not whether the number of watt-hours is accurate or (assuming it is correct) that it's a good amount to spend on each query.

Yep, watt-hours are correct. Think of it this way: Power supplies and laptop chargers are rated in watts and that represents how much energy they can pull from the wall at a point in time. If you wanted to know how much energy a task takes on your computer and you know it maxes out your 300 watt desktop [0] for an entire hour, that would take 300 watt-hours.

[0] To my knowledge, no desktops are built to max out their power supply for an extended time; this is a simplification for illustration purposes.

Agreed. I realize it’s a metaphor but saying we’re so close to the singularity that it’s guaranteed to happen (the event horizon being a point of no return) seems like wild hubris.

It means the shortage is being relieved. From the article:

> The construction boom, nonetheless, has improved rent affordability, according to Realtor.com data.

I still use borg for local backups but use restic for all my offsite backups. Off-hand I don’t think redtic lacks any feature borg has (although there’s probably at least one) after they added compression a few years ago.

I sort of understand the mindset that elaborate notetaking systems are a waste of time because 95 percent or more never gets used but the value of something like Apple Notes for tiny details that are both necessary and relatively tough to remember seems hard to dispute. Do you memorize things like part numbers, confirmation numbers, and dimensions?

(If this is a joke, well, I salute your dry humor.)

Right, my point is that we don't have the data to make a similar exponential argument. We can't rule out the possibility that we're currently in the early stages of exponential growth based on direct measurement. If it is exponential, once it doubles enough times, it will show up in overall economic data.

We can also look at the tools, which have improved relatively quickly but don't appear to be improving exponentially. GPT-4 and GPT-4o came out about a year after their predecessors. Is GPT-4o a bigger leap that GPT-4 was? Are GPT-4.5 or 4.1 a bigger leap than GPT-4 was? I honestly don't know, but the general reception suggests otherwise. The biggest leaps recently seem to be making models that perform roughly as well as past ones but are much smaller. That has advantages from the standpoint of democratization and energy consumption, but those kinds of improvements seems to favor a situation where AI augments workers rather than replaces them.

Yeah, for this analogy to work, we’d have to see AI causing a small but consistently doubling amount of lost jobs.

I’m not sure all of them are LLM generated, but many are. Bizarre indeed.

I’m a little confused; it “looks like Python” but basically every snippet has elements that don’t look like Python. Ignoring the semicolons, which you do call out as a difference, I see let for introducing variables, new for creating objects, and a completely different lambda syntax.

Maybe the “looks like Python” means it uses significant whitespace? In that case I think the README should have an example demonstrating that.

To be clear, you can diverge from Python syntax all you want. I think you just need to be clearer about what you mean by “looks like Python,” maybe by providing an example that highlights the features that particularly resemble Python.

It makes sense to me. After seeing a bunch of AI slop, people started putting no AI buttons and disclaimers. Then some people using AI for little things wanted to clarify it wasn’t AI generated wholesale without falsely claiming AI wasn’t involved at all.

Hard to say from the outside, but it’s possible the service relies on some outdated or insecure library, runtime, service, etc. they want to stop running. Although frankly it seems just as possible it’s a trivial expense and they’re cutting it because it’s still a net expense, goodwill and past promises be dammed.

Static HTML and caching aren't special cases by any means, but a message board where literally nothing changes between users certainly seems like a special case, even twenty years ago. You don't need that in order to make a site run fast, of course, but that limitation certainly simplifies things.

Personally, I wouldn’t normally think of a friendship as a “success accelerator” because that reduces a friendship to its material benefits. I think you can have strong and meaningful relationships without this aspect at all.

Still, I do want to acknowledge that if I’m in a position to help accelerate a friend’s success, I will. I don’t know whether I’d say I expect my friends to do the same for me since it’s so contextual. At the same time I would expect a certain level of reciprocity if they’re in the position to reciprocate, although I admit I’m struggling to articulate my expectations clearly, since they’re not simply one to one.