klabb3

18 ms·

211.
by klabb3 1y ago
> Programmers aren't deterministic either. No but programs are. An LLM can be a programmer too, but it’s not a program the way we want and expect programs to behave: deterministically. Even if a programmer could perform a TLS handsh
212.
by klabb3 1y ago
You’re right about the trees but wrong (hear me out) about the forest. Yes, programming isn’t always deterministic, not just due to the leftpad API endpoint being down, but by design - you can’t deterministically tell which button the user
213.
by klabb3 1y ago
> but in my experience as a tech entrepreneur in Europe, if you want investment, you talk to the Americans Yeah, but that’s not so much cultural but rather the US capital markets are fueled by foreign demand of USD. The same inflated cap
214.
by klabb3 1y ago
That's news to me. I thought companies’ decision making is governed by proxy of the shareholders, not employees.
215.
by klabb3 1y ago
LLMs don’t own data centers nor can they be registered to pay taxes. This projection is not a serious threat. Some would even say it’s a distraction from the very real and imminent dangers of centralized commercial AI: Because you’re right
216.
by klabb3 1y ago
Thanks. This is pretty damn cool, and sounds much more useful than random for real-world use cases. Question: does this work for interpreted languages? Or is this an assembly only thing?
217.
by klabb3 1y ago
Right. Great idea. Maybe call it ”formal execution spec for LLM reference” or something. It could even be versioned in some kind of distributed merkle tree.
218.
by klabb3 1y ago
> I think that if you give the same task to three different developers you'll get three different implementations. Yes, but if you want them to be compatible you need to define a protocol and conformance test suite. This is way more
219.
by klabb3 1y ago
> Most code doesn't express subtle logic paths. If I test if a million inputs are correctly sorted, I've probably implemented the sorter correctly. This just rings of famous last words to me. There are many errors that pass thi
220.
by klabb3 1y ago
My prediction is that the general limitation of multi-step agents is the quality of the reward function. If you think of LLMs as throwing shit at the wall and see if it sticks, but unlike traditional bruteforce (which is ”random” in the out
221.
by klabb3 1y ago
I agree. That’s an interesting model for chat. But remember, modern email is quite complex when you add all the standards for authenticity and security together, as well as informal requirements like spam filtering. But even if self-hosting
222.
by klabb3 1y ago
Yup. Non-traveling US Americans mostly won’t understand how critical WhatsApp is in many parts of the world, for more than a decade. It’s much much stronger than the iMessage norm in the US. Businesses put WhatsApp numbers on their stores,
223.
by klabb3 1y ago
No system can credibly claim to sustain massive consumer scale until proven in real life. Even if your technical architecture supports scale and federation, these are just some threats off the top of my head: - spam, fraud and Sybil attacks
224.
by klabb3 1y ago
This appears to me like the NATS ”request-response” pattern. They also have first-class support for this in their client libs. Under the hood, they create and subscribe to an ephemeral topic where servers can send the response to. (Perhaps
225.
by klabb3 1y ago
This link 404s for me. Is the repo public?
226.
by klabb3 1y ago
> I will likely have my own personal self-driving vehicle. The self-driving car ”utopia” (or rather moderate improvement) very much hinges on the space savings on roadways and parking, to increase utilization, reduce congestion and allow
227.
by klabb3 1y ago
> 4 : Second Order Selves (e.g. cat). Where access consciousness begins. Theory of mind. Self-awareness. Inner narrative. Anticipating the reactions of predator or prey, or navigating a social hierarchy. Cats and dogs most definitely ant
228.
by klabb3 1y ago
> React native can do it, no? In theory yes? But the point is still to use shared API and models for 2-3 platforms. In practice it could go unused if it’s too different. Most real world app developers don’t like to spend maintenance cycl
229.
by klabb3 1y ago
> and are able to do complex manouvres in the air If there’s something pilots really don’t like, it’s when someone else does things without them knowing exactly what, when and why. There’s a reason you separate pilot flying from pilot mo
230.
by klabb3 1y ago
> It also seems like a way to try and go directly against things like React Native and Kotlin Multiplatform. This is a bingo for me. If you know your web, you know these effects are almost impossible to pull off. Or any other UI framewor
231.
by klabb3 1y ago
> Bureaucracy at it finest I truly don’t get this. Is it a Europe = bureaucracy take? > they came up with IP's that can't be more random Came up with? I imagine those novelty IPs are extremely difficult and expensive to acqu
232.
by klabb3 1y ago
> we (i.e. the Jemalloc team) weren’t really in a great place to respond to all the random GitHub issues people would file Why not? I mean this is complete drive-by comment, so please correct me, but there was a fully staffed team at Met
233.
by klabb3 1y ago
> does it mean that the average ride length is 1.8km? that seems kinda low..., like that's something I would walk if I didn't hurry.. Idk about the average but I used to make a bad joke that walking is considered an extreme spo
234.
by klabb3 1y ago
And carsickness. In stop-sign city traffic, I get nauseous with the breaking and speeding of aggressive driving. I mean stop signs are problematic for other reasons too, but I don’t want to get to a dinner with friends feeling sick. That sa
235.
by klabb3 1y ago
To add to this: I’ve never been gaslighted more convincingly than by an LLM, ever. The arguments they make look so convincing. They can even naturally address specific questions and counter-arguments, while being completely wrong. This is p
236.
by klabb3 1y ago
That’s exactly what I was thinking when I read the article. ”Why are you thinking this is meant to work?”. But of course, normal trusting people will believe it if it’s on the ”trusted app stores”. These apps with graphs are just selling yo
237.
by klabb3 1y ago
I propose a slightly different boundary: not ”to self-host” but ”ability to self-host”. It simply means that you can if you want to, but you can let someone else host it. This is a lot more inclusive, both to those who are less technical an
238.
by klabb3 1y ago
Revenue only covers money flowing in. Virtually all companies in an extreme hype race like this one is deliberately losing money total (ie expenses are higher than revenue). Even during shit-hype times like food delivery apps, the companies
239.
by klabb3 1y ago
> I fund $1000 in Claude Code Opus 4 for my top developers over the course of this month, and I really do expect to get >$1000 worth of more work output. Probably scales to $1000/dev before we hit diminishing returns. Two unknown
240.
by klabb3 1y ago
> Mozilla should Why is HN obsessed with suggesting strategic decisions for this company in particular? It’s like the most popular thing to have an opinion about, and only on HN.

More ›