Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
chrisjj
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
12 ms
·
361.
▲
by
chrisjj
2mo ago
What surprises me is anyone is surprised. Obviously a stocastic parrot has no understanding, so any conveyance of true understanding it delivers will be rare and accidental.
362.
▲
by
chrisjj
2mo ago
With no explanation, how will you know whether it genuinely is proof?
363.
▲
by
chrisjj
2mo ago
Then with no way to prove the explanations, anyone can be a "mathematician".
364.
▲
by
chrisjj
2mo ago
> Since people started using Claude now I get full context of everything… Really? How does Claude gather that context?
365.
▲
by
chrisjj
2mo ago
> A short reply from you beats a long one from a model, every time. Disagreed. Bad writing from a chatbot beats worse writing from its user every time - regardless of length. One reason we have so much bot slop is the attractiveness of b
366.
▲
by
chrisjj
2mo ago
Very clear. Thanks. So does your engine automatically work out that the VIEW door blocks the PIC doorway?
367.
▲
by
chrisjj
2mo ago
And script is what sets it to walkable?
368.
▲
by
chrisjj
2mo ago
Great. And the elevator door is operated by script?
369.
▲
by
chrisjj
2mo ago
> Room art (PIC/VIEW) and obstacle polygons are read too, since some gates are geometric and exist nowhere in the script. Anyone got an example? Not counting bugs such as the floor gap in the SQ2 start room.
370.
▲
by
chrisjj
2mo ago
The Readme is actual work.
371.
▲
by
chrisjj
2mo ago
We don't even know what those immediate tasks are. And given the evident spectacular ineptitude of their keepers, I doubt they can be trusted to know either. We could be one prompt injection attack away from internet-wide catastrophe.
372.
▲
by
chrisjj
2mo ago
Distillation is a thing.
373.
▲
by
chrisjj
2mo ago
> Can't a lot happen within ~60 minutes? Spawn a ton of unpausable processes, I'd say.
374.
▲
by
chrisjj
2mo ago
This works on only people dumb enough to block only their own name, right?
375.
▲
by
chrisjj
2mo ago
As a collective we make no choices.
376.
▲
by
chrisjj
2mo ago
> Zuckerberg promises an agent that understands you, your goals, everything you care about. I.e. the most dangerous program you can imagine.
377.
▲
by
chrisjj
2mo ago
Did I read this right? The objects are still just surfaces, not solids?
378.
▲
by
chrisjj
2mo ago
> Because malware breaking the sandbox of one of them can mostly only spy on the owner, while malware breaking the sandbox of the other can spy on whoever the owner walks past, GPS trace their movements… Then fix your s****y sandbox, App
379.
▲
by
chrisjj
2mo ago
> I can't believe the EU Commission agreed to it. This press release conspicuously lacks any suggestion that it did agree.
380.
▲
by
chrisjj
2mo ago
> elevated errors English too difficult for you, Dario?
381.
▲
by
chrisjj
2mo ago
And now the "AI" industry is doubling down on it - with hatbot prompt injection.
382.
▲
by
chrisjj
2mo ago
I'm interested to hear what aspects of the original promise you'd say remain unfulfilled.
383.
▲
by
chrisjj
2mo ago
I'm interested to hear what aspects of the original promise you believe remain unfulfilled.
384.
▲
by
chrisjj
2mo ago
> Nothing in the PR jumps out as a red flag. Made by AI?
385.
▲
by
chrisjj
2mo ago
> a single quote in the title breaks out of echo '...' and allows arbitrary command execution. Quote injection still alive and well in 2026. Gawd.
386.
▲
by
chrisjj
2mo ago
Donkey, surely.
387.
▲
by
chrisjj
2mo ago
No. It means they have a capacity problem.
388.
▲
by
chrisjj
2mo ago
They've had a couple of months and still have not brought in sufficient extra capacity?? What's their excuse?
389.
▲
by
chrisjj
2mo ago
> Do they even tell you what the material is? They do. Click the link.
390.
▲
by
chrisjj
2mo ago
You are not the target audience. The "AI" label is intended to attract stupid people.
More ›