Jon Olson

I build AI supercomputers @google. Amateur circuit maker (and breaker); sometimes skiier. Opinions are my own.

[Jon Olson @jonolson]

@rakyll Bit of a broken record here, but...

Can confirm, living the dream 😁.

[Jon Olson @jonolson]

@rakyll Can confirm, living the dream 😀

[Jon Olson @jonolson]

@rakyll Especially if they're RAM-heavy -- damn.

[Jon Olson @jonolson]

@tqbf Yes, but...

[Jon Olson @jonolson]

@gruber I ask AIs for Markdown regularly because, as you note, it's what you start with. Basically, going back to the inputs, which is fantastic for feeding into downstream AIs at relatively low token cost.

It's a shame that (almost) nobody just shares the starting Markdown.

[Ethan Mollick @emollick]

As working with AI agents looks more like management, we may want to consider large-scale management training for the AI era. The US government actually did this once, & the WW2 Engineering, Science, and Management War Training program was a large reason for the post-war boom.

[Jon Olson @jonolson]

In the interest of trying to answer my own question, I asked some friendly AIs. Their consensus view seems to align with yours (although lots of curious little disagreements between them). They seem to view this as a primary-only play.

Gemini https://t.co/ssK6RdiP4F
Claude https://t.co/E16ZGkQ0Vz
Chat https://t.co/T3h4NMgvPw

[Jon Olson @jonolson]

@NateSilver538 Is there / have you and Eli analyzed any polling on how the Bidens are faring with the broad electorate at this point post-presidency?

I know the common narrative, and that narrative is intuitive. I'm curious what data has been collected, and whether the data aligns with it.

[Jon Olson @jonolson]

@TracketPacer Many of the people that I share this with are going to be deeply enraged, and that makes me happy 😁

V.Important Q, though, what's the chocolate bar of choice for OSFP instead of QSFP?

[Jon Olson @jonolson]

We're going to need Mickens to give us an update to The Night Watch. We're going from

I HAVE NO TOOLS BECAUSE I'VE DESTROYED MY TOOLS WITH MY TOOLS.

to

I HAVE NO TOOLS BECAUSE MY TOOLS HAVE DESTROYED THEIR TOOLS WITH THEIR TOOLS.

No comment on how much closer things brings us to "RIVERS OF BLOOD ALL DAY EVERY DAY."

[Jon Olson @jonolson]

@rakyll ...and for my next trick, reinventing Kubernetes as a simple agentic loop!

[Jon Olson @jonolson]

@puf @the_thagomizer The hardbacks are intended to be ordered chronologically, based on my best interpretation of the relationships between Cryptonomicom, Anathem, FALL, TS, and Snow Crash (TDA is clearly post-SC).

That said, Seveneves feels very out of place.

@nealstephenson thoughts? 😅

[Jon Olson @jonolson]

@puf @the_thagomizer Early years for him, but it's a good one.

[Jon Olson @jonolson]

@TracketPacer Now though, absolutely, yes!

[Jon Olson @jonolson]

@TracketPacer This wouldn't have rung true for more until the last year, when I learned how many upper-stack workload failures happen in certain software stacks in the face of one misbehaving L1 electrical channel.

[Jon Olson @jonolson]

@kellabyte There's a complicated story here, including major Kubernetes as a Service providers not using said storage solution behind the scenes.

[Jon Olson @jonolson]

I was never a heavy user of @OpenAI (Gemini caught up quickly enough), but I had been paying them for a "Plus" subscription.

I've stopped that.

[Jon Olson @jonolson]

@RideoutTim You may want to look into the availability of @HyperinflationTim on various social networks if you want to continue that line.

Should be easy to grab?

[Thomas H. Ptacek @tqbf]

Whoah, this is well said. Haven't seen filtering golfed down that tight before.

[Google Research @GoogleResearch]

Introducing ATLAS: New scaling laws for massively multilingual language models. We offer practical, data-driven guidance to balance data mix and model size, helping global developers better serve billions of non-English speakers. Learn more: https://t.co/8FsHLBKsou