usaar333

6 ms·

31.
Series A startup Equity is worth 4x the preferred price(amafinance.org)
4 points by usaar333 1y ago|0 comments
32.
by usaar333 1y ago
An employee can repeat jobs over and over. Assumption is an exit eventually occurs - same thing VCs feel as well.
33.
by usaar333 1y ago
I didn't claim my risk is lower, but as my link notes I can quit and recall my investment while the investors cannot.
34.
by usaar333 1y ago
Startup equity is worth a lot: https://www.amafinance.org/startup_comp/
35.
by usaar333 1y ago
Not everything is adversarial. More cash pressure on the company itself can be bad for the company which is bad for you too. I always take more equity. I wouldn't work for you in the first place if I didn't believe in your equity
36.
by usaar333 1y ago
Sure, if you are single with no family and wiling to live outside California.
37.
by usaar333 1y ago
Under any normal circumstance I've ever seen, you should be taking the higher equity/lower salary combination and should focus on equity rather than salary. The only time it ever makes sense to push for more salary instead is if y
38.
by usaar333 1y ago
Odd take. Openai gives 5 months of paternity leave and author is independently wealthy. What difference does it make between spending more time with a 4 month old vs a 4 year old? Or is your prescription that people should just be retiring
39.
by usaar333 1y ago
At best about half. Compared to UK, roughly other half of gap is drug overdoses, car crashes and firearm deaths https://publichealth.jhu.edu/2024/americans-die-younger-than...
40.
by usaar333 1y ago
They don't seem to use any recent top models. No opus, no o3, no Gemini 25 pro
41.
by usaar333 1y ago
When you do a chat are reasoning traces for prior model outputs in the LLM context?
42.
by usaar333 1y ago
Why? AI beat rainbolt 1.5 years ago: https://www.npr.org/2023/12/19/1219984002/artificial-intelli... AI tends to have superhuman pattern matching abilities with enough data
43.
by usaar333 1y ago
Been true since gpt-4.
44.
by usaar333 1y ago
> Neither of us will ever pay for power again. Seems optimistic? With 2 EVs and almost all electric utilities, I'm well past the ~26 KWH a day electricity your system might generate. I'm also impressed how cheap your costs are.
45.
by usaar333 1y ago
Huh? It's not even that high on the leaderboard: https://geobench.org/
46.
by usaar333 1y ago
> Some sources mention that o3 scores 63.8 on SWE-bench, while Gemini 2.5 Pro scores 69.1. It's the opposite. o3 scores higher
47.
by usaar333 2y ago
Or OpenAI is? After using Gemini 2.5, I did not feel "AI is slowing down". It's just this model isn't SOTA.
48.
by usaar333 2y ago
They have. https://x.com/arcprize/status/1905361753678246308
49.
by usaar333 2y ago
VIX is still at 35%. We are still in quite unstable territory, just not insanely unstable.
50.
by usaar333 2y ago
> Highest score was 5%, supposedly SOTA last year was IMO silver level. No LLM last year got silver. Deepmind had a highly specialized AI system earning that
51.
by usaar333 2y ago
And then within a week, Gemini 2.5 was tested and got 25%. Point is AI is getting stronger. And this only suggested LLMs aren't trained well to write formal math proofs, which is true.
52.
by usaar333 2y ago
Ya, I find this hard to imagine aging well. Gemini 2.5 solved (at least much better than) multiple real world systems questions I've had in the past that other models could not. Its visual reasoning also jumped significantly on charts
53.
by usaar333 2y ago
I think so. Thane is aligned with the high p doom folks. 1 year may be slightly exaggerated, but it aligns with his view
54.
by usaar333 2y ago
More ability to kill everyone. That's harder to do with nukes. That said, the actual forecast odds on metaculus are pretty similar for nuclear and AI catastrophies: https://possibleworldstree.com/
55.
by usaar333 2y ago
Author also made a highly upvoted and controversial comment about o3 in the same vein that's worth reading: https://www.lesswrong.com/posts/Ao4enANjWNsYiSFqc/o3?comment... Oh course lesswrong, being heavily A
56.
by usaar333 2y ago
LW isn't a place that cares about credentialism. He has tons of links for the objective statements. You either accept the interpretation or you don't.
57.
by usaar333 2y ago
This reddit thread covers it: https://www.reddit.com/r/mlscaling/comments/1izubn4/gpt45_vs... Seems on par with what expected, but there's lots of unknowns
58.
by usaar333 2y ago
For OpenAI perhaps? Sonnet 3.7 without extended thinking is quite strong. Swe-bench scores tie o3
59.
by usaar333 2y ago
Updated. #1 with thinking
60.
by usaar333 2y ago
> Nowhere in the prompt they specified it shouldn’t cheat I'm dubious that in the messy real world, humans will be able to enumerate every single possible misaligned action in a prompt.

More ›