thefourthchime

I find it a little irritating when somebody literally copies and pastes a bot response without editing it at all. Not because it's wrong, but because it's full of cruft and slop. The core information is there, but it's surrounded by junk the author probably doesn't understand and hopes I will.

As a software engineer talking to other software engineers, I'm the controller of my domain knowledge. But my domain knowledge is now subjugated to Claude, and it tells me what it did. When someone else needs to know what I did, I have to ask Claude what it did. Then I send it over the fence and the other engineer does the same.

I don't think this is inherently bad. It's sort of like two managers talking to each other and having to talk to their employees about what's actually going on. What's bad is when the managers themselves don't understand it and just start blindly copying and pasting things. That's where you get into trouble.

Very cool idea!

Best. comment. ever.

I was just thinking, I bet if I refresh hacker news, a new model will come up.

Because it's entirely possible that Western labs already did this optimization but didn't publish it, and then the Chinese figured it out and decided to brag about it.

We don't know either way, so I find the whole thing silly to speculate on.

The image and text selection is somewhat bizarre.

There's one that says, “I just got married. How do I change my last name?” Photo of woman in the back of a car wearing a wedding dress, makes sense.

But -- “Help me find a job” shows a Black man alone on an escalator.

Then, “How do I invest in my child's future?” shows a kid face‑planted on a frozen lake.

“How do I replace my social security card?” has an old man on the beach hunting for metal.

“I'm a veteran. How do we get care?” shows a guy just randomly standing in the back of his pickup truck.

One thing I will say positive about this website is it doesn't look like it was made by a large bureaucracy over many years. It looks like it was made by maybe three people over a week.

I love this because, it reminds me that when people hate AI, They hate low effort AI slop.

This website demonstrates a tremendous amount of human effort, and it walks through the iterative design process that led to the final result. Taking inspiration, trying out ideas, tinkering—that's design.

I'm late with Pac-Man as well..

GPT 6.1 Sol — 91, ~9 min, $0.51 https://jonclegg.github.io/pacman-bakeoff/#gpt-6.1-sol

Opus 5.5 — 99, ~9 min, $2.00 https://jonclegg.github.io/pacman-bakeoff/#claude-opus-5-5

GPT 6 Astra — 87, ~10 min, $2.42 https://jonclegg.github.io/pacman-bakeoff/#gpt-6-astra

Opus still plays the best. Sol is almost as good and way cheaper. Astra costs the most, scores the least of the three, and the UI is full of slop copy and design.

Full gallery: https://jonclegg.github.io/pacman-bakeoff/

The move is only creepy if the interest is one way.

This is why the memes of some guy talking to a coworker is considered sexual assault, where if Tom Brady does it, it’s considered cute flirtation

I don't think the science is settled, but the idea that there's a lot more to do at home to stay occupied and entertained rather than going outside is factually true.

When I was growing up, there were only a handful of channels and then cable. Compared to what we have now, that was the Dark Ages. There's so much you can do on a screen now that's compelling.

Infinite scrolling on Shorts, YouTube, video games, social media, hanging out with people online, chatrooms—combine that with Uber Eats, and you can do whatever you want.

The one thing that will make you want to go out is, of course, meeting someone to have sex with, but that's become weaponized, especially for young adults.

Boys being called creepy and predatory for making the wrong move was always an issue, but now you can be shamed online with a video as proof. Why take the risk when you can meet somebody online?

The game mechanics are shockingly hard for them to get right. You also reveal some of their personality traits. The GPT‑6 and Astra models tend to include a lot of cringe copy as well as their own personal graphics style.

The first thing I did with computers that had any success was a WWIV BBS in the LA area called Altered States.

I think I ran it on a 386 under MS‑DOS. At some point I was getting about forty calls a day and the line was busy a lot. I would drop into sysop mode and talk to people sometimes. One of those times was actually a girl in real life. She ended up being my girlfriend for a couple of years -- fun times....

Yes, Opus 5.5 is the first one that I don't have any notes on. It's just as impressive on other tasks I've given it over the last week. A clear step change.

I need to add that one! It's the first one I've seen to have almost a drunk Pac-Man. It lags a little bit.

It was something I randomly tested about a year ago, and no models handled it well, so whenever a new model came out, I tried it.

It has the advantage of taking good screenshots and letting me know if it’s decent within five seconds of playing.

I'm afraid those will be too easy. I'm not sure what the next game should be...

Just added! I had Opus 5.5 look at them, not a perfect way to score them but it's close-ish -- Best would be a ELO, where people play both and rank a winner, but I don't know if people want to bother doing that.

Thanks!

Your welcome!

Yes! The point being that up until yesterday, every model struggled with this, and now they don't.

It does vey well at one shotting a PacMan clone, pretty much perfect. https://jonclegg.github.io/pacman-bakeoff/entries/claude-son...

2nd only to Opus 5.5, which is perfect. https://jonclegg.github.io/pacman-bakeoff/entries/claude-opu...

Up until very recently, all models struggled with this.

All results: https://jonclegg.github.io/pacman-bakeoff/

Exactly!

On a Lark, I asked Codex to find silhouettes for all car models so I could make a fun drag coefficient website for all cars.

It found a website that had all of them but had no interest in making them available. So it went ahead and started hacking CAPTCHAs and downloading them. I was pretty flabbergasted that it would do this, but also kind of amazed. Eventually I stopped it because I realized I didn't want to be caught stealing these things.

This was around April, the same time as these hacks.

I disagree. Toyota is one of the only companies that didn't plunge into EVs and get completely destroyed. Instead, they've stayed on the hybrid train, which has turned out to be a highlight for ICE cars.

I don't get it... Is this just some kind of meta skill? You should sync up with that other HN site that turns a website into a product demo : )

I have to say so far Opus 5.5 gives me much better results than Fable 5.1 did. I can actually understand what it's talking about.

I’m a grok user, but 4.7 is just worse and more expensive than Opus 5.5 or 6 sol.

Do you know the political positions of all the CEOs you purchase from?

Sigh, do you know all the political positions of all the CEOs you purchase from?

The Nazi thing is so dumb and old.