ngruhn

4 ms·

31.
Lean Explained with TypeScript(gruhn.me)
2 points by ngruhn 2mo ago|0 comments
32.
by ngruhn 2mo 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
33.
by ngruhn 2mo ago
agree, despite the em dashes.
34.
by ngruhn 2mo ago
Tell me about. Well it should have MCP access in plan mode to lookup backstage docs, right? . Agent proceeds to launch playwright sessions...
35.
by ngruhn 2mo 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
36.
A Test for AI Consciousness(gruhn.me)
1 points by ngruhn 2mo ago|1 comments
37.
by ngruhn 2mo 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
38.
by ngruhn 2mo 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
39.
by ngruhn 2mo 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
40.
by ngruhn 2mo ago
Oh he's left handed? Ma man!
41.
by ngruhn 2mo 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.
42.
by ngruhn 2mo ago
Not OP but maybe those are Bitmapped Vector Tries: https://www.infoq.com/presentations/Functional-Data-Structur...
43.
by ngruhn 2mo 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
44.
by ngruhn 2mo ago
Author here. I have indeed not.
45.
by ngruhn 2mo ago
I've been hearing forever that latent reasoning is the next step. But none of the frontier labs seem to incorporate it. Any idea why? It doesn't sound so technically challenging. Sure you loose some interpretability but surely eve
46.
by ngruhn 2mo ago
Fair enough but cutting the tax is another incentive measure: make the legal way to buy cigarettes the cheapest way. As the article states: the fact illicit tobacco is illegal does not prevent it and poses massive enforcement costs.
47.
by ngruhn 2mo ago
Another example: In Germany the chancellor decided that workers take too much "sick leave". Fair enough (probably) but they decided to clamp down on it by requiring a doctor's certificate from day one. Rookie mistake! For the
48.
by ngruhn 2mo ago
I know it's kinda beside the point but I like incentives compared to hard rules as a policy measure. They are much more effective, less bureaucratic and easier to push through. They are like weights vs. booleans. Don't like smokin
49.
by ngruhn 2mo ago
I don't disagree but for some of the examples given (YouTube, Instagram, X.com) a huge entry barrier is the network effect. Even if AI could perfectly clone these products in a day, that still doesn't translate to instant market s
50.
by ngruhn 3mo ago
Genuinely open to that. What's weak about it?
51.
Don't be a meat proxy(gruhn.me)
1849 points by ngruhn 3mo ago|741 comments
52.
by ngruhn 3mo ago
Ramen is a great soup. But I haven't been blown away in any shop not even Japan. EXCEPT that random place in Kyoto we went into without plan. I think it's even a franchise. Not an ancient shop with 100 year tradition or something:
53.
by ngruhn 3mo ago
Not many though
54.
by ngruhn 3mo ago
As always there is also the gradual sunset option: slap a tax on it and increase over time. - That's easier to get past the lobbies. - It still surfaces unexpected consequences but reduces impact.
55.
by ngruhn 3mo ago
But everything that's different seems like an arbitrary requirement. I hope I'm also conscious in moments where I'm not actively growing/killing neural connections.
56.
by ngruhn 3mo ago
The level of conspiracy thinking he here is reaching COVID levels.
57.
by ngruhn 3mo ago
> I don't see why AI company PR statements are relevant here They suppress the news --> proof of malicious intent. They disclose the news --> proof of malicious intent.
58.
by ngruhn 3mo ago
How would you want them to behave? Suppress the news?
59.
by ngruhn 3mo ago
That's a beautifully simple characterization. I've never heard of of that. What keywords do I need to find more on this topic?
60.
by ngruhn 4mo ago
> “Safety” in the context of LLMs concerns not offending certain individuals. It absolutely concerns not destroying the world.

More ›