Every week I learn I’m a n00b when it comes to leveling up with Al.
I’m hiring for facilities maintenance, and specifically seeking a high-agency self-starter who can help us optimize bathroom breaks. An internal candidate sent over multiple Claude Code generated iterations of the 💩emoji. One of them had blonde hair and Madonna’s famous cone bra.
I was speechless.
Who wants to join me in standing up the competing Center For Applied Rationality Techniques
guest writes for Forbes would produce.
I seriously think we can completely dismiss Forbes as a credible source at this point, even if it’s not something coming from, ahem, “contributors”
Oh don’t worry, bespoke GLP-1s plus his “nootropic stack” have surely helped him out immensely by now
And now that we’ve got Quantum within AI, just imagine everything you can haul within Quantum
Meanwhile, “AI agents” continue to be an opaque bundle of shell scripts shoved into a trenchcoat, with an inconsistent English-language translation layer stapled on top
Setting firm boundaries is for amateurs, haven’t you heard?
Is this… technocracy in collapse…
Do they also block NYT, WSJ, Atlantic? I think not. Odd, but I didn’t trust dang and friends to start with. HN is not intended to be a neutral forum.
the Hofstadter tendency towards neopronouns
Seeing it extracted from context and called out like this helps me understand why I’ve bounced off Hofstadter multiple times over the years, despite his hype. It’s an artistic choice, sure, but 400 pages of this stuff without a break can be like beating your face against a brick wall after a while.
centaur-configured programmer
Cory, baby, my dogg, sure “enshittification” was a big hit, but you can’t expect that your rough-draft followups are automatically gold
you gotta give him a morsel of credit, he’s got his buzzword and he’s stickin’ to it
“Aging left” has lost “vitality” - he’s phoning this one in, straight out of the house style guide.
Hmm, he’s still sticking to tweet-threads on Twitter. We’ll know he’s fully cracking when he resorts to Ackman-style unreadable text blocks on there.
is now completely one shotted.
I’m not sure “one-shotted” is a good description for some of these folks. More like they bet against themselves in a rigged game of Russian roulette being played with a fully-loaded Uzi.
man, I got totally zonted in the zonte last time I was down there
yeah, you know, I’m really doing OK with just the one kidney. seriously. thanks for asking
As a layperson skimming the paper, this strikes me as equivalent to a dashed-off letter to the editor coming from someone in Knuth’s position. It’s an incomplete, second-hand reporting of somebody else’s results that doesn’t really investigate any of the interesting features of the system at hand. The implicit claim (here and elsewhere) is that we have a runtime for natural-language programming in English, and the main method reported for demonstrating this is the partial prompt:
** After EVERY exploreXX.py run, IMMEDIATELY update this file [plan.md] before doing anything else. ** No exceptions. Do not start the next exploration until the previous one is documented here.
and later on, a slightly longer prompt from a correspondent using GPT-5.2 Pro, that also loads a PDF of Knuth’s article into the context window. No discussion of debugging how these systems arrive at their output, or programmatically constraining them for more targeted output in their broader vector space. Just more of the braindead prompting-and-hoping approach, which eventually, unsurprisingly diverges from outputting any viable code whatsoever. This all strikes me as being an exercise similar to
You are a cute little puppy dog. Do not shit on the floor. Do not deposit bodily waste or fecal matter onto hardwood, linoleum, tile, and especially not carpet. Do not defecate indoors. Do not consume your own fecal matter.
The cargo-cult system prompt approach is like banging two rocks together compared to what a computational system should be capable of, and I would be much more impressed and much more interested if someone like Knuth was investigating such capabilities, instead of blogging somebody else pretending to have the Star Trek computer.
“They wanted me to build an AI, so I built a shoddy AI casing filled with used pinball machine parts!”
tit-for-tat escalation
I think maybe you nailed it here. Being able to pretend they’re doing game theory/mapping out an escalation ladder allows Yud and our friends to feel like they’re in the same intellectual lineage as guys like Oppenheimer and Teller, manifesting the same sort of “objective” emotionless rationality. The big difference, you see, is that the AI will think so much faster than us that…!
“Look, Grandpa, here’s what we should invest the trust fund into!”