@ryiacy

applied AI for GTM teams, DMs open. ☧

 london
Joined June 2021
  • Tweets6.6K
  • Following2.3K
  • Followers3.1K
753Media
😂
109
2,207
102
21,080
527,338
My AI agent is vagueposting in my WhatsApp DMs again
2
33
967
I will return to this paragraph by C. S. Lewis at least once a week until the day I die
100
5,105
99
35,308
678,411
increasingly wondering if the whole notion that software needs rebuilding for agents via becoming headless might be completely wrong and it's just gonna be that they're gonna interact via UIs and computer use
10
1
1
33
1,429
Computer use agents are the humanoid robots of software agents. They're not the 'first principles' solution to anything. They're the 'worse is better' quick fix solution. But they play nicely with everything and the world is built for humanoids
2
8
133
i basically am cheering on openai because i appreciate altman's honest sociopathy and will to power and prefer these to the anthropic program for enriching berkeley cultists
143
143
23
3,394
114,888
OpenAI and Cursor now seem to be switching positions. OpenAI have the best developer tooling for building agents (OSS Codex + now Agents SDK built on top of it) while Cursor have the best agent for non-technical people in Grok Bot
2
9
445
so many failed attempts at building new useful devtools for AI agents (most obviously langchain). it seems like what's actually useful is just better quality devtools in general, since agents are developers
4
9
693
In anything coding adjacent, a lot of AI engineering techniques seem to now be technical debt and building multi-step workflows is an antipattern that you should replace with a long-running sandbox agent. In others, e.g. anything writing related, it still requires lot of handholding
3
11
1,191
The key question for whether pacing is the frontier is legitimate or cartel behaviour seems to be: is it about slowing down development, or slowing down releases? if the labs are willing to slow down development, i.e. make slower progress on R&D, that's clearly legitimate. If it's to continue to make rapid progress internally but not release models to the public, it's less obvious
3
6
652
Clearly to some extent it's both but right now the labs want to have their cake and eat it. If there is information that they have internally that indicates they should slow down, they should transparently share it, not insist that they be allowed to maintain secrecy
65
Gonna tell my kids this was Cognition's pitch deck
Michael Morelli
@morellifit
Sep 7
The most commonly rated “life-changing” interventions: - Doing math - More sleep - Weightlifting - α-Klotho - Oxytocin - Dihexa - Modafinil - Nicotine - Caffeine Personally, I’d add bromantane and semax to the upper-right quadrant too.
Readers added context they thought people might want to know
This chart is a digitally altered version of a chart from a 2022 Reddit nootropics rating survey. It has added numerous nonsensical nootropics and has repositioned real nootropics that were included in the original survey. web.archive.org/web/2026031300…
1
16
980
I do respect the fact that highly talented individuals with math PhDs want to spend all their time studying math simply for love of the game. But it seems probably net good for society if those individuals have to get highly paid and interesting quantitative roles in industry instead (maybe not if they all became hedge fund quants, but now there's clearly massive demand and great comp for AI researchers, SWEs, etc) So it just doesn't seem like a massive issue to me if math gets automated away. Plenty of us have passions that we can't monetize, usually with worse fallback options!
4
16
767
I'm confused that people are confused by the fact that AI doesn't shows up in GDP statistics yet. In order for that to happen, businesses need to deploy it, and there's a massive amount of work to be done to make that happen which hasn't been done yet
3
1
18
905
I found out today that French people say "wouaw" instead of "wow". They have an endearing attachment to maximizing number of vowels in words
1
12
361
It's so strange that many rationalists are completely incapable of communicating with any gravitas. If you believe this, then you have a responsibility to talk about it seriously, not in this dysregulated exclamation-point heavy EA style
Evan Hubinger
@EvanHub
Sep 9
Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to.
12
4
2
423
16,110
GPT Astra, which finished pre-training in July, estimates the probability that an AI would solve a Millenium Prize problem in 2026 at 3%
3
1
1
84
2,979
This is much more optimistic than Fable, which in July estimated the probability of an AI solving the Jacobian Conjecture at 0.1% nitter.cf/ryiacy/status/20791924…
dom
@ryiacy
Jul 20
Claude Fable estimates the probability of an LLM solving the Jacobian Conjecture this year as <0.1%
1
2
485
As AI makes our lives and work more and more comfortable and easy, the ability to "stare into the abyss" and pay attention to things that are uncomfortable to think about is going to become more and more important
3
4
43
1,014
one of the best blog posts, from @benkuhn benkuhn.net/abyss/

Staring into the abyss as a core life skill

thinking about scary things • examples from Wave • examples from elsewhere • finding a buddy • getting the timing right • a list of abyss questions

benkuhn.net
5
111
Load more