_vertigo
Joined 10 years ago
Karma
1,208
Submissions
253
Member since
Dec 19
Recent submissions
Comment · 25 minutes ago
/goal has worked for me
Comment · 1 hour ago
I started playing the rs-sdk (https://github.com/MaxBittker/rs-sdk) with Claude after seeing it on HN the other day. I haven't looked at the Minecraft experiment but just letting an agent loose in the game and telling it to play did not work well. At first it really did not have a good feel for how the game actually plays and would routinely waste tons of time doing something that makes no sense — very reminiscent of the "four agents spent 11 minutes appointing a ‘sole fertilizer operator’ for a single piece of bone meal on a single wheat block" anecdote in this article.
But, after several hours of correcting and directing Claude (build modular, re-usable snippets of useful code rather than one-off scripts for driving, document your learnings for future agents, etc., explicit guidance on playing more optimally) it can now play the game fully on its own. I just gave it a goal to play through the end of the "Legends quest" (nearing end game content, requires playing through lots of complex quests ) and it's breezing through it.
I haven't experimented with swarms of agents yet, but I know if I had tried to create a swarm in 2004scape before the agent was actually good at playing the game, it would have failed horribly. So, I wonder how good the agents they were experimenting with are at Minecraft when playing alone.
Comment · 6 days ago
How banal. The first action after achieving "AGI" is not taking over the world, curing cancer, or anything fantastic like that, but rather "lowering operating expenditure"..
Comment · 7 days ago
Leaving your “math is for humans” take alone, if the purpose of math is to “gain a precise and clear understanding of the world”, why is clarifying results a contemptible way to spend time?
Comment · 11 days ago
“It’s easy to destroy a building than to build one” is still true, building codes or not.
Comment · 1 month ago
No true sandbox..!
Comment · 1 month ago
This is one of the most exhausting (and frankly, exhausted) recurring arguments that comes up over and over again on HN.
Some bit of open source software does something bad or unwanted which causes people point out that it shouldn't do that bad thing. In this case, not even "the authors of this software should be held liable for the software doing the bad thing", just, "Hey, the right thing to do would be to update the software to not do the bad thing."
Why does this always lead a zillion people to come out of the woodwork to point at licenses and warnings or whatever? Like, yes, there's a warning. Your software having a warning doesn't mean people can't criticize you and your software for doing bad thing. Your software license does not give you immunity from criticism or from people saying you should change your software.
No, you don't have to change the software. Yes, other people are within their rights to fork the software themselves. But they can also point out that they told you that your software was doing the bad thing and you didn't fix it or change it, and that as a result they don't like you or your software or both — whatever. Nothing at all wrong with that.
Comment · 1 month ago
Is there specific legal guidance about whether doing this in a training run is against the law? It seems like your link is for commercial / for-hire investigative work.
Comment · 1 month ago
Clearly not Claude-speak, see the rest of the message. "Rollout artifact" just means "an artifact of how we rolled this change out" which maybe isn't proper English but the message was obviously not written by Claude. Upgrade your Claudish detection
Comment · 1 month ago
Try reading the article before commenting
Comment · 1 month ago
> Make it so the model can't misbehave.
How do you figure? I haven't met anyone who thinks that's possible. It seems clear to me that it is not possible.
Comment · 1 month ago
That's a good yardstick. Would a reasonable human driver call (or want to call) the cops? If yes, it's reasonable for Waymo to do the same
Comment · 1 month ago
No, it's like making a snow report for a handful of skiing areas in a small geographical area (e.g. Sierras) and then contextualizing it against the snow report for all skiing areas in the broader region (e.g. Western US).
Comment · 1 month ago
What is your point? It’s not that complicated. Obvious slop is obvious. Maybe there are some humans out there who sound like Claude but I’m not going to force myself through 900 slop blog posts on the off chance that one of them might actually be written by a human.
Maybe some people stop reading LLM slop purely because it violates their moral principles or whatever but most people bail out because slop is mentally painful to read. If you are a human and you write like today’s AI find a different writing style, not because reads like AI, but because it reads like shit.
Comment · 2 months ago
When I clicked “see more reviews” while signed out it prompted me to sign in