vessenes

41 ms·

961.
by vessenes 1y ago
This is just so, so wrong. LLMs change the surface of what's "hard" to do in a coding exercise. Many a project has so much boiler plate, edge cases, etc. that months+ can be taken up dealing with what is ultimately a very bor
962.
by vessenes 1y ago
oAI rev exceeds $1b per month right now.
963.
by vessenes 1y ago
iOS 26 public beta is out now; very near public release I’d say. I’m typing to you on it right now
964.
by vessenes 1y ago
To a mathematician saying god created the integers is the same thing as saying god created the rationals: there’s a bijection between the rationals and integers. I’m not convinced that we could have our current universe without irrationals
965.
by vessenes 1y ago
Terrible, terrible leadership at the top of the AI org. Plus a fundamental commitment to being a 'product first' company. This same commitment would be why I wouldn't count them out on the AI side, btw. It's not clear th
966.
by vessenes 1y ago
That's almost certainly available in a public database in China, FYI. I got to speak at a cool event in Shenzhen in 2019 and the gates automatically face (and maybe iris?) recognized for entry -- I never 'got scanned' conscio
967.
by vessenes 1y ago
I mean you could just click "bulk inference rates" and wait for cheap inference time to reify your avatar message to your kids.. I like going back to the 60s where a lot of futuristic scenarios come complete with magnetic tape sto
968.
by vessenes 1y ago
Thank you for this expansion. I was about to rabbit hole on how it could be that ratio-based trig (and what is that?) is more accurate than modern calculations. Re: rationals, I mean there's an infinite number of rationals available ar
969.
by vessenes 1y ago
Cost to develop in today's dollars: $50bn (more if you consider it as a % of GDP). Cost per kg to launch something with it: roughly $70k. Cost to launch on falcon per kg: $2-3k. Wait, that's price. SpaceX is profitable. It's
970.
by vessenes 1y ago
I understand you’re worried about publishing your code. I’d be happy to help do something with this at a larger scale, but I think I could use a litttle more detail. Are you saying the training task is to ask for the (fib_i)th token rather
971.
by vessenes 1y ago
I think it's a human thing. The Torah is succinct; The Talmud has a lot to say about it. For a large codebase, the comments would be huge, and also I think distracting. In fact, as a former code auditor I can say that comments at times
972.
by vessenes 1y ago
And Blue Ant is required reading for modern marketing. Also its awesome
973.
by vessenes 1y ago
Yep he did, according to the top comment
974.
by vessenes 1y ago
Man I can only hear this in Keanu's voice.
975.
by vessenes 1y ago
I think he was generally just excellent at getting near term predictions right and interesting . To pick a small one of many there's a moment in Heavy Weather where they have trouble with their PBX, and he mentions that it's ba
976.
by vessenes 1y ago
Rudy, thanks for putting this up! I have a paper copy in my library, but digital is great, and now the panopticon AIs can admit in public they've read it, rather than secretly torrented it in their training data. Immortality of a sort.
977.
by vessenes 1y ago
Notable here I think is the agent workflow - as LLMs continue to have increased 3d world understanding, they’re going to be useful in a variety of circumstances. Taking the human out of the loop for error checking and bug fixing is going to
978.
by vessenes 1y ago
I want to play! But.. localhost:~# echo -e "\x00XR2K" | nc clearsky.dev 29438 AXR2KPxr2000.leo.spacenet:send a single 0 byte followed by 'XR2K' for documentation printf has the same problem. Is it me?
979.
by vessenes 1y ago
If you want to feel worried, check the Altman AMA on reddit. A lottttt of people have a parasocial relationship with 4o. Not encouraging.
980.
by vessenes 1y ago
Well i vote you as better than either :)
981.
by vessenes 1y ago
Gemini is in hard sycophancy mode here; it knows you want it to take the piss and it’s giving you what you want. Presuming the last two are from 5, they are to my eyes next generation in terms of communication — that’s a spicy take on neuro
982.
by vessenes 1y ago
Hard disagree. The essay is a rehash of Reddit complaints, no direct results from testing and largely about product launch (simultaneous launch to 500mm+ users mind you) snafus. Please. I think most hit pieces like this miss what is actuall
983.
by vessenes 1y ago
After a few hours with gpt-5, I'd trade that spread. Not that I think oAI will win end of year. But I think gpt5 is better than it looks on the benchmark side. It is very very good at something we don't have a lot of benchmarks fo
984.
by vessenes 1y ago
Yessss. I feel this. On the productivity side I banned Reddit and gave myself 15 min of HN a day, that helped with not getting waylaid. On the pleasure side, I can get close to the same feeling with a design and review cycle going while Cla
985.
by vessenes 1y ago
Don’t be mad bro. Seriously. Every single person working on a film has creative input, not just someone hand painting a backdrop. You have an immense number of tools available to be creative with now. This is a great thing!
986.
by vessenes 1y ago
This is super creative. For those who like to reduce their experiences of fun cool stuff, I’ll describe it: Implement a line drawing-on-2D algorithm that maintains geometry on a particular 2d view. Run each of your source images for a Gauss
987.
by vessenes 1y ago
Ghz speed video processing, even if we only get very basic segmentation or recognition out of it, is probably crazy useful. Need to face recognize every seat at a stadium? Well, if you have enough cameras, 60,000 seats could be scanned 250
988.
by vessenes 1y ago
This seems to work in hardware, per the paper. At least to 80% accuracy.
989.
by vessenes 1y ago
To be fair, the most-forbidden technique is a concept and a proposal, not an iron law. I don’t work at Anthropic, but I imagine internally that their “helpful only model” — the model that does not refuse, or the base model —- that model has
990.
by vessenes 1y ago
Actually, Anthropic has put out some research showing that hallucination is a thing their models know they do; similar weights are activated for ‘lying’ and ‘hallucinating’ in the Claude series. Implication - Claude knows - at least mostly

More ›