I was once SUN Microsystems Campus Ambassador at Oxford University and our first lecture was about dtrace! The guy from SUN was blown away by attendance so the higher ups wanted to do more & more events in bigger rooms. But as corporate weeks passed, I started telling them to postpone because everyone was entering the exam sessions. Corporate of course did not listen. Long story short, SUN sent 3 experts but we couldn’t get the crowds and the main lecture room was embarrassingly empty.
Imho 5.5 is the best balanced model that has the highest focus and balanced delivery particularly in web development. It's not great at research or novel development tasks, but for building web apps it’s the best. Sad that OpenAI wants to retire it.
Claude models were always too eager and "overly helpful" for my taste. But it seems better models tend to be this way. GPT 6 and 5.6 are overly helpful too, but at least less than Fable. But I seem to be sticking to GPT 5.5 as this was a really focused model.
You could probably make your infra cheaper if you sent the generated slides to the web client and use some js ffmpeg wasm to compile the video in the browser. This way you could probably use the cheapest VPS to handle this.
"option shift -" like so —, or if you want an en-dash, just "option -" like –. The funny bit is that I like em-dashes too as they indicate a pause or an interruption in thinking that just looks better than parenthesis, however I sometimes like to use en-dashes as well, which can be a pain since some fonts, especially monospaced ones, render exactly the same, for example in this HN edit box: - – —
I used to get 2-3 shady crypto offers per week on LinkedIn. It stopped when I started replying with AI generated responses demanding multiple verification steps: official email, official offer link, terms and scope etc. And a note with a firm refusal to run any code or install any package on my machine for "recruitment tasks".
The PR buzz convinced me so I subscribed today to Pro. Running two tasks simultaneously with Fable and Opus 4-8 on ultra reasoning, analysing a single smart contract file used all my 7h usage within 20mins and didn’t produce any results. Pretty useless. I think Anthropic has plenty of room to optimise the interactions and token use but that would cut their income quite a lot, I doubt there’s any will to do it pre-IPO.
Dosbox has its emulator to run DOS, but afaik dosbox binary that starts up dosbox on Mac uses Rosetta on M-series Macs, and without Rosetta it simply won’t start and emulate DOS.
Hmm i guess this would require figuring out how to extract game data from an .exe installer GOG distributes. I guess AI would figure it out in like 3 minutes ;) I’ll check later.
So I guess sci-fi movies were right all along. Nobody in Star Wars knows how hyperspace travel works, it just works. The little robots know everything but almost no human bothers to care. People just carry on with their bickering lives while the bots whiz in the background, and these robots are astonished at human inefficiency every single time, but rarely do anything about it. And people are still people.
So I guess that yesterday's questions "Where is vibecoded Photoshop?" just got answered with a vibecoded Gimp UI that looks like Photoshop ;) but seriously, nice work!
Well my idea was that when I said that the overall story arch is interesting, one would read a paragraph or two and see themselves if it's worth reading further or not, so I didn't think mentioning anything else is necessary ;)
You're correct these text are much better called short stories, but I chose the word "novel" anyway. And despite the slop, I still find this experiment interesting for one: because it intrigues me how AI ponders on the subject of sovereignty, and for two: because when I tried other, human novels I also encountered slop (albeit human): pages of meaningless descriptions, deliberations and dead-end side-plots.
This read is amazing and the development work is very impressive, great job and congrats! That said, my 20-30yo self would end at that. However, my 40yo+ self has a piece of wisdom here: the brass plugs are there for a reason: they slow things down. Technocracy (screens, apps, automation) is not good for our mental health. Human minds need small, calm, slow, manual processes. Like plugging the brass plugs.
As a kid behind the Iron Curtain I had a radio & cassette tape player with a single large speaker built-in. The cassette player had a record button that when pressed together with play used the speaker and I could actually record my voice, and play it back, from the tape. After enough re-recording on a single tape the fade of the old recordings broke through the new ones creating a truly amazing sound experience. Too bad I lost the tapes and the device.
You learn the most from failure and the least from success. Likewise, you grow the most from pushing through the limits and the least from living in abundance. You can do pretty much anything on a full spec Mac Studio, but so can anyone else with a lot of money. But if you push through the limit of a MacBook Neo, you just did something no one else was able to do. And that is awesome.
What is dead may never die. Perl will live on in the few mostly weird places doing things other languages can only dream of. And that’s fine. Perl doesn’t need to win the race or update itself. It is good enough as it is. It’s the magic wand in the world full of electronics. Sure you can light the room with an electric PHP/Python lightbulb script. But Perl can summon a levitating glowing orb out of thin air with a single line of code.