vessenes

13 ms·

241.
by vessenes 5mo ago
This. The paragraph might have backed up and said "Bezos, after sustaining over 95% capital losses in the prior decade,.."
242.
by vessenes 5mo ago
I don't think this changes the final conclusion - but have you considered calculating against depreciation -- i.e. figuring out how much your M3 ultra is worth today, and only charging yourself for the delta? In my mind you might even
243.
by vessenes 5mo ago
I have some quibbles about the ProPublica definitions -- for instance market liquidity matters when calculating public company stock wealth -- and even if you're going to borrow against it, there are additional costs and pledges that m
244.
by vessenes 5mo ago
I think you may have missed the background: US tax rhetoric -- he's doing what I think is pretty fair math with a fair take -- the math is supposed to break down what percent income tax you need to get the same dollars in tax revenue a
245.
by vessenes 5mo ago
$0 just feels like a concept -- I can imagine a really high quality structuring exercise that gets tax low by making sure leverage on capital is what's used for spending, but I'd be really surprised to see a 0.1%-er (or 0.001%-er)
246.
by vessenes 5mo ago
Nope. Just not post things like "billionaires pay no taxes."
247.
by vessenes 5mo ago
This just isn't true, unless you're the president. Who is the single largest taxpayer in US history? I'll wait while you google it.
248.
by vessenes 5mo ago
Please make higher quality posts -- what in specific do you think pg has missed or does not understand?
249.
by vessenes 5mo ago
There's a related calculation you can do -- what percent of your net worth is your employability? Take your salary, divide by 0.05 (or multiply by 20) -- if you had that much additional wealth earning 5%, you could replace your job&#x2
250.
by vessenes 5mo ago
For sure. There’s also this kind of interesting meta strategy question: if you think you can float the huge capex with some internally valuable processes as a base case then you might just go really big. Dario and Elon’s temperaments come
251.
by vessenes 5mo ago
Thank you. It’s arguably even more tiring to hunt it down on HN, and often thankless.
252.
by vessenes 5mo ago
I mean … they’re mostly sitting on their SpaceX stock and rubbing their hands gleefully. What do they have to keep up with? Disclosure: former (small) SX investor.
253.
by vessenes 5mo ago
Love the curiosity. I recommend you spend that curiosity on figuring out why both companies are in the situation they are that this deal is a win-win: Elon has provided a masterclass in public company launch and management over the last 15
254.
by vessenes 5mo ago
To the contrary - there’s so much demand that Dario needs Elon’s compute and is willing to pay premium month to month pricing for it.
255.
by vessenes 5mo ago
Boy I’d love to have the sort of ‘no real advantage’ that got me $1.9bn in revenue per month from Anthropic. Make no mistake - Elon’s charging premium pricing because he has something Anthropic realllly needs. In exchange, he will get signi
256.
by vessenes 5mo ago
For reference: here’s my post at the time about it: https://news.ycombinator.com/item?id=43732866 — that includes the original trace. This is a fairly heavy ontological argument; in my case, I meant quite simply that it cal
257.
by vessenes 5mo ago
Interesting. Agreed that’s a significant list. Post model 3 launch, Tesla had a number of senior folks leave almost immediately. My read at that time was they had hit or exceeded pareto-optimal on the suffering:wealth scale —- Tesla was cle
258.
by vessenes 5mo ago
When we were all discussing this the first time around, I noticed in the CoT that o3 was cheating — it pulled location data, then lied about it. From recent Anthropic mechinterp work, it looks like models have likely moved lying into their
259.
by vessenes 5mo ago
It might be more useful, but as is, it is still dispositive: 5.5 is significantly worse than o3 at geo-guessing. And the “magic” prompt doesn’t matter that much, at least in o3’s case.
260.
by vessenes 5mo ago
Good q. You can see that tantalizing graph where compute is displayed on a log axis and pass@1 goes up to like 50/60%. I’d love to know just how much compute is encoded on that axis. I guess you can get some estimate from the excerpted
261.
by vessenes 5mo ago
Oh I didn't say they'd be consistent. But once it's raised, it's going to be really tough.
262.
by vessenes 5mo ago
Sort of buried the lede here -- Apple uses the Inkwell name and has a trademark. This is just not going to get approved. Or, to quote Jobs talking to the iPodRip developer "Change your apps name. Not that big of a deal."
263.
by vessenes 5mo ago
This is a very small item, but I found it interesting that the paper does not credit Stability AI in the author bylines.
264.
by vessenes 5mo ago
Interesting. Have you worked in China?
265.
by vessenes 5mo ago
All this data is accessible to national security agencies; this is true in every country in the world. China has more integration between intelligence and industry than many western countries, and it does present a higher risk of unwanted “
266.
by vessenes 5mo ago
Some of it’s probably timing. Some of it is wanting to look good. That said, I just went to the claw-eval site, and neither 4.7 nor 5.5 from oAI are listed on the benchmarks. So there’s also just the time from others to get benchmarking don
267.
by vessenes 5mo ago
Be careful taking the headlines at face value - that list of people leaving was mostly product and redundant senior execs to my eyes, post spacex merger. You’d expect those folks to be asked to leave as part of a re-org in any event. I don’
268.
by vessenes 5mo ago
I agree with that. Right now, you choose: Subtle bugs in understanding the spec but strong arch and coding (codex) Or Subtle bugs in implementation but good understanding of the spec (claude).
269.
by vessenes 5mo ago
Hmm, I read the situation a little differently. Grok is not a slouchy model. It’s not the best, but it’s not the worst. X currently has one source of proprietary data, Twitter, and grok is by far the best at all the things you might imagine
270.
by vessenes 5mo ago
To your point, I didn't even look at the code.. :) Okay, I looked at the codex code. it's super reasonable -- separation of concerns, operating on a state model, it's not over designed. I did not hate it. I also noted that co

More ›