causal

9 ms·

181.
by causal 6mo ago
Unfortunate if so, but I'm finding plenty of counterexamples in the past day alone: https://hn.algolia.com/?dateRange=last24h&page=0&prefix=true...
182.
by causal 6mo ago
"Not aliens" seems obvious but shouldn't be a basis for dismissing this either. I feel like sometimes we are so determined to dismiss aliens that we accept any plausible alternative too quickly, when there might be something
183.
by causal 6mo ago
You seem to be going off the title which is plainly incorrect and not what the paper says. The paper demonstrates HOW different models can learn similar representations due to "data, architecture, optimizer, and tokenizer". "
184.
by causal 6mo ago
Title is editorialized and needs to be fixed; the paper does not say what this title implies, nor is that the title of the paper.
185.
by causal 6mo ago
9 nines found somewhere after the decimal point if you measure with enough precision
186.
by causal 6mo ago
Yeah! It's such a delightful site, I love these sorts of odes to a niche passion I had not yet learned to appreciate.
187.
by causal 6mo ago
This is an interesting topic but reporting on what some random people typed or clicked on social media is such a shallow basis for news. It's a subjective narrative of a subjective trend.
188.
by causal 6mo ago
Once again random tweets from insiders being the only clues we have to what Anthropic actual policy is
189.
by causal 6mo ago
I think it's probably pointing toward the general harm that thinking only in objects has done to programming as a practice
190.
by causal 6mo ago
Exactly. Today "premature optimization" almost always means unnecessary infra or abstractions or some other complexity- not DS&A choices.
191.
by causal 6mo ago
Wildly different scenarios. You don't exactly get to pick how conveniently your enemy aligns themselves with objectives.
192.
by causal 6mo ago
Cool. If I understand correctly though, the single-kernel only works on a single GPU right- no parallelism support to go Q8 on 2x3090?
193.
by causal 6mo ago
> We've also lost the focus on clearly defined objectives for war. Are you saying we had this in Vietnam? And I don't think the evidence is strong that these "hard lessons" did anything to keep that same generation fr
194.
by causal 6mo ago
I like the idea of civic engagement / service in theory too, but I feel like the Vietnam war was a demonstration of possible failure modes when draft is in place: a lot of poor kids died, some rich kids allegedly used parental influenc
195.
by causal 6mo ago
I am planning to cancel my Claude sub after I've used up remaining tokens this weekend. Model is fine but the harness and reliability have absolutely nosedived. Just unacceptable for the high price. Vibe-coded way too much of their own
196.
by causal 6mo ago
Will give it a try but my experience with Claude and browser use so far is that it’s extremely lazy: it rarely notices or cares when something doesn’t look right, it needs lots of pointing out “hey you ignored that broken render” etc
197.
by causal 6mo ago
"Not everything" - sure, but mass surveillance and autonomous killing are kind of big things to sweep under that rug no?
198.
by causal 6mo ago
Output tokens are more expensive
199.
by causal 6mo ago
The email use case is interesting, though I feel like dictating via intermediary AI sounds more tedious to me than just using the email app on my phone.
200.
by causal 6mo ago
Thanks for sharing, can you give any specific examples of the kind of personalized tools that you find helpful?
201.
by causal 6mo ago
100%. Every time I see a Claude Code announcement I sigh when I still can't get the desktop app to fullscreen without fritzing or make a plan and display it consistently. Many papercuts.
202.
by causal 6mo ago
It's like they oversold capacity and are looking for excuses to get rid of users lol Opus is fine but it's not THAT much better than the alternatives. I want to support Anthropic because they seem "less shady" than OpenA
203.
by causal 6mo ago
Claude sets an environment variable to prevent nested invocation, but I've found it can be unset. No idea if they consider that a violation of the secret laws though.
204.
by causal 6mo ago
Haven't Github-triggered LLMs already been the source of multiple prompt injection attacks? Seems bad.
205.
by causal 6mo ago
Yeah in the span of a month or so we had: - SDK that allows you to use OAuth authentication! - Docs updated to say DO NOT USE OAUTH authentication unless authorized! [0] - Anthropic employee Tweeting "That's not what we meant! It&
206.
by causal 6mo ago
World: What's your uptime? Anthropic: 9. World: 9 nines? That's awesome! Anthropic: No, a singular 9. 90-something percent uptime.
207.
by causal 6mo ago
That is something of a broken promise in the SaaS world; it was supposed to be convenience but instead it became a hundred broken integration points as you struggle to keep up with deprecated APIs and acqui-hires killing off shims you would
208.
by causal 6mo ago
Totally agree, though I'd argue that it's still a software failure if preventing exploits requires every user memorize and follow an onerous list of best practices.
209.
by causal 6mo ago
Yes, previous models found vulnerabilities but Mythos is uniquely capable of actually exploiting them: https://red.anthropic.com/2026/mythos-preview/
210.
by causal 6mo ago
No, Opus has found a lot and 112 vulnerabilities were reported to Firefox alone by Opus [0]. But Mythos is uniquely capable of exploiting vulnerabilities, not just finding them. [0] https://red.anthropic.com/2026/mythos

More ›