Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
chrisjj
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
10 ms
·
241.
▲
by
chrisjj
2mo ago
There'd be a huge problem training enough replacement humans to do the work to the low standard of the retired bots. We might have to settle for a rise in the quality of work. /i
242.
▲
by
chrisjj
2mo ago
> A series of highly unlikely conditions that all need to be fulfilled for this to become a vulnerability. Poor thinking. A vulnerability is a vulnerability regardless. The conditions apply to the exploit .
243.
▲
by
chrisjj
2mo ago
> *I mean killing someone by taking control of a system and misusing it resulting in someone's death, not an "indirect" death caused by the providion of incorrect information in a chat app. What's the difference?
244.
▲
by
chrisjj
2mo ago
> But make your program generate a few sentences based on statistics It is even easier. Simply make your program refer to itself as "I". Uniquely amongst your examples, LLMs are powered by human gullibility.
245.
▲
by
chrisjj
2mo ago
Then you missed the word "can".
246.
▲
by
chrisjj
2mo ago
> If the productivity gains from AI are so big, spending one day a week to minimize its downsides shouldn't be a difficult trade-off. That is not minimising.
247.
▲
Everything's broken and nobody's upset (2012)
(hanselman.com)
1 points
by
chrisjj
2mo ago
|
0 comments
248.
▲
by
chrisjj
2mo ago
More significant I think is paying for software indicates you are more selective - inc. against defects.
249.
▲
by
chrisjj
2mo ago
By that token every CSS1 page is bugged.
250.
▲
by
chrisjj
2mo ago
> It is well known that we can read faster with narrower text Speak for yourself. Font size is client setting, and there's where lies any bug.
251.
▲
by
chrisjj
2mo ago
> At what point does search quality degrade to the point where is becomes a bug? It depends entirely on the commitment made by the service, so e.g. my desktop file search program results has bugs whereas Google's has none.
252.
▲
by
chrisjj
2mo ago
> Software that does not meet expectations, especially in a field like search, which is basically a long running war between SEO and search engines, is not a 'bug'. While bug has various definitions, not meeting reasonable exp
253.
▲
by
chrisjj
2mo ago
> Is half of the industry under AI psychosis right now ? A bad coder sees bot code as an inprovement. A good coder sees bot code as deterioration. They are both correct.
254.
▲
by
chrisjj
2mo ago
> But are LLMs actually getting smarter, or just better at fooling us? They look smarter to people they've made dumber.
255.
▲
by
chrisjj
2mo ago
Grift poorly defined? https://www.merriam-webster.com/dictionary/grift
256.
▲
by
chrisjj
2mo ago
> If it doesn't work well then it naturally loses trust Too late. If it doesn't work well then it has already naturally destroyed trust.
257.
▲
by
chrisjj
2mo ago
> Push a button, every icon is now labeled with a key, type that key, counts as a click, done. So, omits the text-labelled buttons?
258.
▲
by
chrisjj
2mo ago
> this self-assessed AI level for contributions [1] to be extremely useful Why would you trust the bot to self-assess truthfully? That's like trusting a spammer to tag his emails as spam.
259.
▲
by
chrisjj
2mo ago
> What does "Responsible" even mean... Quite. Of the two different common meanings, I reckon: "The use of a generative AI tool does not diminish the contributor's responsibility for the work they submit." will be
260.
▲
by
chrisjj
2mo ago
> The Debian Project nevertheless expects that all contributions submitted to Debian, regardless of how and with which tools they were produced, satisfy the same standards of quality, correctness, maintainability, and legal compliance. H
261.
▲
by
chrisjj
2mo ago
Here Claude takes a long time - fetching 19 webpages - but does then answer.
262.
▲
by
chrisjj
2mo ago
"Auto Mode is a convenience feature backed by a best-effort classifier, not a security guarantee,” Dario, again you've mistaken your best for good enough.
263.
▲
Claude Code can be tricked simply by asking it to summarize a website
(theregister.com)
4 points
by
chrisjj
2mo ago
|
6 comments
264.
▲
by
chrisjj
2mo ago
True title: Researcher shows how Claude Code can be tricked simply by asking it to summarize a website Anthropic’s Claude Code running Opus 5 in Auto Mode can be tricked into executing attacker-controlled code simply by asking the coding a
265.
▲
by
chrisjj
2mo ago
Rendering in 2D and summing per-pixel Z-spans. For a finely meshed shape and approx. result, it has by far the lowest cost.
266.
▲
by
chrisjj
2mo ago
> move to the country side, and homestead So, work from home, but you'll need a new home. What will even pay for it? Not the proceeds from your old home which now nobody wants to buy.
267.
▲
by
chrisjj
2mo ago
> It's OK to offer an app, but if a site is pushing, pressuring or forcing you to get one, it's almost certainly because they want to do something abusive that they currently can't. Every UK bank is, IME.
268.
▲
by
chrisjj
2mo ago
What complaint?
269.
▲
by
chrisjj
2mo ago
I can see why it didn't catch on. The phone user would prefer to use the on-machine UI needed for regular users and bigger.
270.
▲
by
chrisjj
2mo ago
"you must use a computer web browser". Yeah, and I have one on my Android phone!
More ›