Stelios @SteliosPapageo7

Tufts Applied Math | AI compute, quant trading | football, pool, drinks & meeting new people Boston, MA Joined December 2022

Tweets

10
Followers

13
Following

217
Likes

30

Stelios @SteliosPapageo7

17 hours ago

@reinerpope I thoroughly enjoyed both of your Dwarkesh Patel podcast/blackboard lectures. I was wondering if there are any other resources you’d recommend for building a deeper understanding of frontier AI in general, especially the full compute-to-AI pipeline, for an interested outsider like me?

0 0 0 6 0

View Details

Stelios @SteliosPapageo7

2 days ago

@PeterDiamandis @DaveBlundin @Link_Ventures I'm an avid listener of the podcast and really enjoy your thinking around AI and how much easier it may (or has) become to build companies. One thing I keep wondering about is the physical infrastructure underneath all of this. Do we actually have enough generation and transmission capacity to support the amount of compute we are heading toward? I'd love to hear you bring someone onto the podcast who deeply understands the grid and transmission, maybe someone like Rob Gramlich. I ask because I'm genuinely fascinated by this problem and would love to understand the space well enough to build in it one day. What actually needs to be built to power the AI revolution?

0 0 1 14 0

View Details

Stelios @SteliosPapageo7

2 days ago

AI compute is starting to look less like “buying GPUs” and more like aircraft or commercial real estate finance. You have a capital-intensive asset, contracted cash flows, residual value risk, lenders, SPVs, and now banks transferring credit exposure. ft.com/content/9cb639…

0 0 0 16 0

View Details

Stelios @SteliosPapageo7

3 days ago

The interesting question to me is less whether attention can be measured and more how it gets priced. The index gives you a measurement, but the dollar price ultimately comes from counterparties willing to take opposite exposure. The real unlock will be whether there are enough natural hedgers and speculators to create deep liquidity around that signal. If there are, this starts to look less like “betting on virality” and more like a genuine market for cultural risk. pretty cool project though icl.

Owen Botkin @owenbotkin

4 days ago

x.com/i/article/2091…

50 13 125 38K 49

1 0 2 67 0

View Details

Stelios @SteliosPapageo7

4 days ago

@AnfieldPapers @davidornstein can't understand how people don't see this is a parody account honestly and sit there and get aggravated in the comments

2 0 2 9K 1

View Details

Stelios @SteliosPapageo7

4 days ago

@usr_bin_roygbiv $300k with “little or no experience” is doing a lot of work here lol. SF pays well, but this is a pretty exaggerated picture of the market.

0 0 3 218 0

View Details

Stelios @SteliosPapageo7

4 days ago

@CounterPressers really good first impression from him

0 0 0 699 0

View Details

Stelios @SteliosPapageo7

4 days ago

I'm spending the next year trying to understand AI compute from first principles. Not just GPUs technically, but pricing, utilization, capacity, financing, derivatives and how compute becomes a tradable asset. I'll post what I learn here.

0 0 1 35 0

View Details

Stelios @SteliosPapageo7

4 days ago

This paper makes the “idle consumer compute” thesis feel much less hypothetical. A 284B MoE running on one RTX 5090 by using GPU + CPU + RAM together. We may already have huge amounts of AI compute sitting idle in people’s homes. arxiv.org/abs/2608.16157

0 0 0 36 0

View Details

Stelios @SteliosPapageo7

5 days ago

@NOTDNFR @Math_files Putting Ramanujan on the Putnam is risky. The Putnam requires incredibly strict, step-by-step rigorous proofs. Ramanujan was a genius of intuition who rarely wrote out proofs, so he would lose a ton of points on grading criteria alone, even if his answers were brilliant.