@gfodor@fchollet A few exemplary sources I pulled while refreshing my memory:
@GavinSBaker Aug '24 interview
https://t.co/njS2go9vzk
Leo dropped Situational Awareness in June 2024
https://t.co/nOflkpxgCs
See the third footnote. This was 2023
https://t.co/B04ivyqiBC
@gfodor@fchollet There was at least a year where the narrative was that scale would lead to a digital god that could do and capture nearly all economic value for whoever got there first
I remember b/c I had a few existential crises during that time
@dadiomov From you list I've read Vogel's book and McGee's.
This book gives you a bridge from Deng to Xi and is a lens to look at Apple's naivety that was on display in Apple in China
Haven't read your others so I'm not sure if it overlaps or is better represented by other recs
@DKThomp@milansingh03 Halberstam is goated, like a tier below Caro. The Coldest Winter shows some of the other forces that enabled a talent-less loser like McCarthy to be such a menace until he was finally censured in '54 and then kindly removed himself from civilization in '57
@NateJoseph19@cbokhove I've viewed you as a good literacy/meta-analysis source. Your math piece, where I am well versed, I was nonplussed to the point of wondering if it was all Gell-Mann Amnesia. I don't say this b/c I'm in any particular math camp or as a dunk but it & this thread have me questioning
@jennalaib Do you find Ts at your school are able to do, review, then update lesson plans based on cool downs regularly? Plainly, are they able to do this when they don't have you in there as extra support?
Btw, love this and love that you are the eternal flame of #mtbos
@ryanlpeterman It looks like you're getting guests who are close enough to the action to know ball but not so high-status or polished that we end up with a story flattened into corporate mythology. It's a big world and I've always found it useful to get a look behind the curtain
@ryanlpeterman I stumbled on this episode of your podcast and really enjoyed it. Bryan has always been a great interview but you are crafting something here. It reminds me of the best of the early Internet History Podcast
@EldicLeonard@HarryStebbings Seed European founders > Project Europe founders get U.S. VC validation > Puts competitive pressure on European VC ecosystem for better terms & speed
Saw what @giffmana has been saying: Codex > Claude. One prompt, Codex built a working version https://t.co/WSoEE5ksDD
Claude was left self-reflecting โWhat I learned is that I have a failure mode where I hit resistance, pivot to an easier solution, and then rationalize it"
@manaltdan All of that stuff models do in their CoT that takes time make real-time frontier STS hard.
Could also be data, compute, lab priorities, or other challenges.
But youโre the one at Deepmind so disabuse me of any incorrect notions here
@manaltdan Latency
Iโve been building speech capability into an app and STS is all smaller models that sacrifice ability for low latency.
If you want frontier capability, gotta use STT > Model > TTS and deal with the latency.
@GregKamradt On days like that I do a heavy hold first. Lights the CNS up and makes the working weight feel better.
You ever try them?
https://t.co/A95uGN5HvN
@lennysan I get the metrics, but your best interviews are those with people who are close enough to the action to have a good view but far enough removed to have a less polished narrative about it. AI/tech exec interviews are high noise, low signal slop
@surmenok@ID_AA_Carmack I enjoyed 1632 and 1633. Then they started publishing side quests with multiple authors. The Galileo Affair was ok. I gave up at the Ram Rebellion. The main thread got picked back up after these so maybe that's worth continuing to read