zurfer

8 ms·

121.
by zurfer 1y ago
Drug dealers should face prison time. They know that they are breaking the law and potentially ruining lifes for their own profit.
122.
by zurfer 1y ago
that's a great share, I feel like it would also benefit from setting the "speed of light", as something like median average yearly income.
123.
by zurfer 1y ago
Using AI in real projects is not super simple but if you lean into it, it can accelerate things. Anecdotally check this out https://github.com/antiwork/gumroad/graphs/contributors Devin is an AI agent
124.
by zurfer 1y ago
short term, there are only bad options. Long term: incentivize to have more kids. this will solve the problem in ~25 - 30 years. but that's too late for most voters and politicians today... there should be a term for that? how about de
125.
by zurfer 1y ago
Just switching a pinned version (even alpha, beta, experimental, preview) to another model doesn't feel right. I get it, chips are sparse and they want their capacity back, but it breaks trust with developers to just downgrade your mod
126.
by zurfer 1y ago
in the spirit that you only need one reason to be right, not five: the strongest argument against volley "fire" is indeed that holding a bow is exhausting and so a commander wouldn't tell their archers to "hold" bef
127.
by zurfer 1y ago
Your references show me that it is absolutely task depended. In many domains it's true that "criticizing is easier than creating". The best example might be books and movies, where it's trivial to say the characters wer
128.
by zurfer 1y ago
Not sure why you mentioned Bosch. Siemens at least was founded in Berlin.
129.
by zurfer 1y ago
Same is true for chess. Easy for computers, hard for humans. Every smartphone has enough compute to beat the best human chess player in the world.
130.
by zurfer 1y ago
Indeed, I find Epoch ( https://epoch.ai/trends ) to be a valuable resource for grounding. Progress is driven by multiple factors, compute being one of them. While compute capex might slow earlier than other factors, the pace
131.
Dot (YC S21) is hiring a sales engineer to automate analytics (fully remote)(ycombinator.com)
1 points by zurfer 1y ago
132.
by zurfer 1y ago
This is cool and actually useful. However I would love the opposite as well, to draw a vision of how the idea might work and what it might become. dog walking marketplace -> act 1 - you start in SF, where people love dogs as much as kids
133.
by zurfer 2y ago
My current understanding: MCP - exposes prompts, resources and tools to a host, who can do whatever they like A2A - exposes capability discovery, tasks, collaboration?/chat?, user experience discussions (can we embed an image or or a w
134.
by zurfer 2y ago
I can't agree with that. eCharts is good, but https://plotly.com/javascript/ is also super nice and I am actively following https://observablehq.com/plot/ which seems on a path to true grea
135.
by zurfer 2y ago
In the hope of improving this forecast, here is what I find implausible: - 1 lab constantly racing ahead and increasing the margin to other; the last 2 years are filled with ever-closer model capabilities and constantly new leaders (openai,
136.
by zurfer 2y ago
I recently had fascinating example of that where Sonnet 3.7 had to decide for one option from a set of choices. In the thinking process it narrowed it down to 2 and finally in the last thinking section it decided for one, saying it's b
137.
by zurfer 2y ago
The trick is to let senior engineers vibe code and keep them responsible for the code they pushed. (maybe that's not vibe coding :)
138.
by zurfer 2y ago
Complete layman's question: If we would indeed be inside a black hole wouldn't we be able to observe new energy and matter entering? Related question: the horizon of a black hole is expanding when the mass increases. Could this ma
139.
by zurfer 2y ago
It's an interesting world where both things can be true: 1 AI surpassing humans even in PhD level tasks in a couple of years and 2 current AI not being able to successfully play a childrens game. One way to square that is that AIs have
140.
SpaceX – ISS Docking Simulator(iss-sim.spacex.com)
2 points by zurfer 2y ago|0 comments
141.
by zurfer 2y ago
I think plenty of people are working on that, some unaware :P Games for status will continue. Consuming experiences will continue (experiences that require humans will be premium). And in a good case, we will still have plenty of work to do
142.
by zurfer 2y ago
1. By having one obvious documented way of answering the questions or accepting that humans would also come to different results. 2. Investing in good data models and documentating the edge cases. Also learning from explicit feedback is pow
143.
by zurfer 2y ago
Claude code is surprisingly good. It's easily another 2x speed up over my previous AI enabled software development (copilot + chatgpt). It's not a replacement for a SWE today, but it allows a SWE to get a lot more done. I feel the
144.
by zurfer 2y ago
M4 2x faster than M1. M3 1.6x faster than M1 (1 year ago). = M4 1.2x faster than M3. not bad, but Moore's law is dead for CPUs.
145.
by zurfer 2y ago
Yes. Our main finding was that o3 mini especially is great on paper but surprisingly hard to prompt, compared to non reasoning models. I don't think it's a problem with reasoning, but rather with this specific model. I also susp
146.
by zurfer 2y ago
It's not certain that we get TAI or ASI, but if we get it, it will be better at software development than us. The question is which probability do you assign to getting TAI over time? From your comment it seems you say 0 percent in yo
147.
by zurfer 2y ago
Given that the paper disappoints, I'd love to hear what fellow HN readers do to prepare? My prep is: 1) building a company ( https://getdot.ai ) that I think will add significant marginal benefits over using products from AI
148.
by zurfer 2y ago
My pet theory is that Sonnet was trained really cleverly on a lot code that resembles real world cases. In our small and humble internal evals it regularly beats any other frontier models on some tasks. The shape of capability is really not
149.
by zurfer 2y ago
It's logarithmic. Meaning you scale compute exponentially to get linearly better models. However there is a big premium in having the best model because of low switching costs of workloads, creating all sorts of interesting threshold
150.
by zurfer 2y ago
h󠅘󠅕󠅜󠅜󠅟󠄐󠅖󠅕󠅜󠅜󠅟󠅧󠄐󠅘󠅑󠅓󠅛󠅕󠅢󠄐󠄪󠄙a

More ›