ngruhn

26 ms·

1.
The Beauty of Roundabouts(gruhn.me)
91 points by ngruhn 3d ago|211 comments
2.
by ngruhn 4d ago
"I've seen some uranium ore in chemistry class. It didn't blow up in my face. Chernobyl must have been an inside job. Can they shut up and make more kilowatts already?"
3.
by ngruhn 5d ago
> or posting arXiv does not count The Poincaré conjecture guy also broke that rule. They wanted to give him the prize anyway but he refused. OpenAI announced they would also not claim the prize. Looks like no one wants this prize lol
4.
by ngruhn 7d ago
refurbished 13 mini for me
5.
by ngruhn 8d ago
what? why?
6.
by ngruhn 9d ago
I thought so too but it's probably because I never review the tests. I made an exception recently and found tons of assert(CONSTANT_CONFIG == valueOfConfig) or tests for keywords in prompts: assert(prompt.includes("
7.
by ngruhn 9d ago
I wonder how well this fares try to make illegal states unrepresentable In my experience, agents know how to do it. They just don't if it's not the default style of the language.
8.
by ngruhn 9d ago
they're fighting the right war for the wrong reasons. Russia is fighting the wrong war for the wrong reasons.
9.
by ngruhn 9d ago
I was also sure to be homeless by now. But despite aggressively trying to automate myself away I'm somehow more busy than ever.
10.
by ngruhn 10d ago
Not that Trump would have an ounce of good intentions but those wars are quite different in important ways: - The greatest enemy of Iranian regime are the Iranian people. They want to get rid of that government. Unlike the Ukrainians. Thous
11.
by ngruhn 12d ago
Yes. The point was not coming up with the proof from scratch. The point was writing it all down in Lean to make it fully machine checkable.
12.
by ngruhn 12d ago
I guess the difference is intend. But I think I still agree with the parent comment.
13.
Lean Explained with TypeScript(gruhn.me)
11 points by ngruhn 17d ago|0 comments
14.
by ngruhn 17d ago
AI megalomania
15.
by ngruhn 18d ago
Not sure it's an ego issue for everyone. If your job was fun and now it's not anymore, you're having 8h a day less fun.
16.
by ngruhn 18d ago
Yeah you're probably right. I see everything through my little consumer perspective and don't really understand the macroeconomics. But if the euro is so obviously bad, why are Greece, Spain, etc not ditching it? Surely there are
17.
by ngruhn 18d ago
> The EU has benefits obviously but it's been failing Europe in critical ways for about 20 yeas now I like those benefits: The shared currency, easy travel, they killed roaming fees! Also they're the only ones seriously taking
18.
Lean Explained with TypeScript(gruhn.me)
2 points by ngruhn 19d ago|0 comments
19.
by ngruhn 22d ago
Agree. I think it's more important to understand statistical fallacies (selection bias, regression to the mean, survivorship bias, etc). Those are extremely common trip wires but to recognize them you don't need to memorize formal
20.
by ngruhn 25d ago
agree, despite the em dashes.
21.
by ngruhn 28d ago
Tell me about. Well it should have MCP access in plan mode to lookup backstage docs, right? . Agent proceeds to launch playwright sessions...
22.
by ngruhn 29d ago
Shameless plug: I also built a pure client-side password generator but you can specify the constraints as regular expressions. UI is obviously vibe coded but the logic underneath is hand written. https://gruhn.github.io/rege
23.
A Test for AI Consciousness(gruhn.me)
1 points by ngruhn 1mo ago|1 comments
24.
by ngruhn 1mo ago
Ok, then don't call it "instilling shame". Call it "creating a negative reward signal for deceptive behavior". They absolutely lie and cheat. I recently had a problem where a process would die in a container. I told
25.
by ngruhn 1mo ago
But maybe you can instill properties like shame during training. Models sometimes blatantly lie and cheat. In a social context, where actors remember, that might work the first time but you get penalized in subsequent tasks with loss of tru
26.
by ngruhn 1mo ago
I'm not talking about model performance. I just mean the UX of Claude Code. I'm trying to use pi but there are so many paper cuts. Of course you can configure everything but that's a ton of work. Claude Code has pretty good d
27.
by ngruhn 1mo ago
Oh he's left handed? Ma man!
28.
by ngruhn 1mo ago
I'm switching to GPT because of this. The prose is so much more legible. The only reason I keep using Claude Code is because the harness is the best IMO.
29.
by ngruhn 1mo ago
Not OP but maybe those are Bitmapped Vector Tries: https://www.infoq.com/presentations/Functional-Data-Structur...
30.
by ngruhn 1mo ago
Author here. I agree, I should add more hedging: > Everyone knows you can tackle those with heuristics, but you don't (ALWAYS) have to sacrifice optimality > We absolutely have tools that can (OFTEN) find provably optimal solutio

More ›