joegibbs

11 ms·

211.
by joegibbs 1y ago
In the Atlantic?
212.
by joegibbs 1y ago
It's in Cursor as well, I've been using it today and I can't say I'm that impressed with it. Definitely not Sonnet 4 or GPT5-high level. It's fast for sure, but it seems to pay no heed at all to instructions. I tell
213.
by joegibbs 1y ago
I think they'll always need special guidance for things like business logic. They'll never know exactly what it is that you're building and why, what the end goal of the project is without you telling them. Architectural stuf
214.
by joegibbs 1y ago
You could have a society where there's one single spreadsheet package made by a team of 20 people, a few operating systems, a new set of 50 video games every year (with graphics that are good enough but nothing groundbreaking so they&#
215.
by joegibbs 1y ago
It would be difficult to do with a raw model, but a two-step method in a chat interface would work - first the model suggests the URLs, tool call to fetch them and return the actual text of the pages, then the response can be based on that.
216.
by joegibbs 1y ago
Yeah exactly, it’s that the text the model is trained on considers poorly-written code to be on the same axis as other things considered negative like supporting Hitler or killing people. You could make a model trained on synthetic data tha
217.
by joegibbs 1y ago
Yeah me too, I think 3 does a much better job for photos or even just images that look like realistic renders. I use 3 for generating grids of age-progressed portraits for a game and it does a better job at sticking to the prompt. 4 also se
218.
by joegibbs 1y ago
I don't think that it's related to any kind of underlying truth though, just the biases of the culture that created the text the model is trained on. If the Nazis had somehow won WW2 and gone on to create LLMs, then the model woul
219.
by joegibbs 1y ago
Amazon has 1.5m, it's more on the level of a small country. Google is only at 180k, just small city level.
220.
Thoughts on GPT-5(jgibbs.dev)
1 points by joegibbs 1y ago|1 comments
221.
by joegibbs 1y ago
Zoom in on Kinshasa though and you can see that it's almost entirely very densely-packed slums and shantytowns built up against each other, about 100 - 250 square meters in size, with no gardens or back yards. Slums can be very dense,
222.
by joegibbs 1y ago
When writing React: - Default export a React component. - Use Tailwind for styling, no import needed. - All NPM libraries are available to use. - Use shadcn/ui for basic components (eg. `import { Card, CardC
223.
by joegibbs 1y ago
I'm using GPT5 on Cursor and it honestly doesn't feel all that much better compared to Claude Sonnet 4 - it's definitely slower as well. I still find it making errors at about the same rate, maybe 10% better max. It's fr
224.
by joegibbs 1y ago
"If you covered a backhoe with skin, made its bucket look like a hand, painted eyes on its chassis, and made it play a sound like “hnngghhh!” whenever it lifted something heavy, then we’d start wondering whether there’s a ghost inside
225.
by joegibbs 1y ago
What if we just banned passive investing? Surely it's bad for the economy to funnel all the money toward the biggest firms rather than the most competitive. If they're performing badly then their stock should crash.
226.
by joegibbs 1y ago
The environmental stuff is so overblown. Speaking of veganism? Eat a steak and that’s 10,000 litres per kilo! Combine that with the methane etc and you’d offset your year’s AI usage. Same with driving your car to the shops. Take the scenic
227.
by joegibbs 1y ago
I think don’t worry too much about trying to avoid it. Think 5-10 years ahead max, rather than 20. In 20-30 years there’s a good chance that what you’ve written will be obsolete regardless - even if programs from 1995 ran perfectly on moder
228.
Show HN: An Infinite Wiki Simulator(wikisim.jgibbs.dev)
3 points by joegibbs 1y ago|0 comments
229.
by joegibbs 1y ago
Nobody would downvote you on Reddit for talking about those topics, unless you’re all in favour of wealth transfer to the billionaires etc
230.
by joegibbs 1y ago
It’s definitely a render, look how clear the small text is, you can see the lower resolution ground texture in the foreground.
231.
by joegibbs 1y ago
It’s the latest in a series of homogenising inventions - the printing press, radio, television, the internet - that will probably result in people of the future speaking and thinking more similarly than today. First went minor languages, th
232.
by joegibbs 1y ago
But given that companies are made up of people, where does the line get drawn? If your concreting business has a couple of other guys working does it still count?
233.
by joegibbs 1y ago
But say you're a concreter, you've worked with a client who's an absolute prick - doesn't do anything illegal exactly but he's a total stickler with your work, calls you all the time with stupid questions, tries to
234.
by joegibbs 1y ago
You could train it to respond in the third person with no point-of-view, like a Wikipedia article, so that it never even refers to itself as "I".
235.
by joegibbs 1y ago
I disagree that Tasmania is economically successful or has a competitive advantage. Out of all major states and territories Tasmania has the lowest GSP per capita, the second-lowest GSP overall and below-average economic growth for the coun
236.
by joegibbs 1y ago
I think it would be good to combine traditional OCR with an LLM to fix up mistakes and add diagram representations - LLMs have the problem of just inventing plausible-sounding text if it can't read it, which is worse than just garblin
237.
by joegibbs 1y ago
It’s really terrible, not only does it get a ton of things obviously wrong but it also gets contradicted by the first search result, and it smugly tells you you’re wrong in an authoritative tone as if you’re some kind of idiot for even aski
238.
by joegibbs 1y ago
If you wanted to you could make the LLM return entirely deterministic results, but it wouldn’t be very helpful since a semantically identical prompt could still create an entirely different result with a single character difference.
239.
by joegibbs 1y ago
For public services, every year people get older, more of the economy has to be reallocated towards looking after them. More spending on pensions rather than education, more old people using all the health services (since they use up so muc
240.
by joegibbs 1y ago
When things go wrong with AI people seem to often make it prostrate itself, explain why it went wrong and promise to never do it again - which it does, but there’s no point, it’s not going to remember because it doesn’t have memory, and the

More ›