techno yoda

bios are boring

[techno yoda @techno_yoda]

@mattpocockuk We need more deeper integration with os level constructs too for things like identity and access control

[techno yoda @techno_yoda]

https://t.co/IvCVuA7ZzV

so damn good

[techno yoda @techno_yoda]

@paularambles in a world where Claude does everything, we should be talking to our coworkers more

[techno yoda @techno_yoda]

@deedydas Any serious security startup would go start hacking websites left right center and patching them as a measure of good will to avoid other hackers screwing them over.

Instant virality and very Silicon Valley coded ( as forgiveness over permission )

[techno yoda @techno_yoda]

@willcb if we go by the idea of just do the simple dumb thing at scale then just make the context ridiculously bigger and train for it to learn in context as well.

[techno yoda @techno_yoda]

@Sirupsen yet

[techno yoda @techno_yoda]

@ElloSunsh1ne @DarioAmodei @_NathanCalvin weaponizing my favorite fictional ai like this for commenting on the cosmic irony of reading an ai policy thread that is so long that most people will treat it like the works of Wittgenstein is something that makes me have very mixed feelings. well done

[techno yoda @techno_yoda]

@AntithesisHQ baldurs gate 3

[techno yoda @techno_yoda]

@Rabadash2 the fact that they decided to invest in cuda when no one in the world was thinking about it clearly shows he knows nothing more than making toys.

[techno yoda @techno_yoda]

@jtimsuggs reminds me of road rash

[techno yoda @techno_yoda]

@nateberkopec u and I meet very different devs

[techno yoda @techno_yoda]

the models just wanna learn

[Nick @nickcammarata]

interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had

with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works

https://t.co/zHij5zwKGT

[techno yoda @techno_yoda]

@maxcembalest @fleetingbits if model quality just gets better then at some point a really hyper efficient AI company would have a way higher portion on r&d since the models would juice out a hyper efficient inference system for whatever they train ( making the inference system comes from r&d budget )

[techno yoda @techno_yoda]

@spicey_lemonade this research direction is Lindy.

[techno yoda @techno_yoda]

sf exploration 📸 (forever expanding)

[techno yoda @techno_yoda]

@nehadesaraju simple napkin math proves this before anyone saying it

[techno yoda @techno_yoda]

@sporadica this is very bell curve meme coded.

[techno yoda @techno_yoda]

@nikillinit https://t.co/WF4eTGR6RK

[techno yoda @techno_yoda]

suddenly everything in that world that was positive sum became zero sum and everything undecidable become positive sum.

[techno yoda @techno_yoda]

suddenly everything in that world that was positive sum became zero sum and everything undecidable become positive sum.