Ray Song retweeted
Agents can now access Computable GPU Index (CGI), the first open source GPU price index via @findatasets.
We're excited to see more players building with Computable!
Introducing our GPU Prices API
See what it costs to rent an H100, H200, B200, or B300 for an hour, right now.
Includes:
• Price updates
• Price spreads
• Daily changes
Agents can now track the price of compute like any other commodity via REST or MCP.
In partnership with @getcomputable
Ray Song retweeted
Always a fan of @WPReadingClub , had a great time chatting about compute market structure at Stanford for SBC 2026 alongside @sui414 and @Raysong
Whitepaper Reading at SBC 2026 finished - 5 sessions across market structure, consensus, and compute markets.
A huge thank you to our session leads for bringing the research and guiding the discussions:
- @xxxin001 (PhD Student at @Stanford)
- @MaxResnick (Lead Economist at @Anza_xyz)
- @MohammadMJalal1 (Researcher at @category_xyz)
- Ryan Kim (Co-founder of @fractalyze_io)
- @0xfishylosopher (Junior Partner at @PanteraCapital )
And thank you to @Stanford and @initc3org for hosting, and to everyone who joined us to read, discuss, and challenge the ideas.
More from each session below🧵
getting a lot of asks from ai companies on how to hedge their gpu price risks through cash settlement derivatives. internally, we’ve built tools to better understand their needs but companies end up spending way too much time playing with this tool so we’ve decided to just bring it to everyone
Ant over LaMelo was clearly the right choice but it’s disappointing that a professional sports team would get nervous and the media would mock a teenager just for dreaming big
LaMelo Ball reportedly scared the Timberwolves away during the draft process when he said he wanted to be the President of the United States, per @JonKrawczynski
“So, we know they picked Edwards. But that’s not my little nugget for you all. It’s why they picked Edwards. There were a million different reasons in their evaluation. But one of them was that they had an interview with LaMelo Ball during this process—trying to figure out, ‘Hey, is this the guy for us?’ And LaMelo… did not interview very well. One of the questions they asked him was: ‘Hey LaMelo, after you’re done playing, what do you see yourself doing? What kind of things do you aspire to?’ And LaMelo said: ‘I want to be President of the United States.’”
(Via hoopshype.com/story/sports/n…)
edge compounds:
better infra/model and more capital → higher win rate and risk tolerance → tighter quotes, more flow, and more market share (very winner-take-all) → more capital to reinvest in infra/model
leaders pull away until microstructure changes or new products appear
I am a long-time admirer of @davidasinclair’s longevity research
also found it morbidly funny that D.A. Sinclair in @InvincibleHQ is a mad biochem scientist who believes humans are flawed and need fixing
wish it were true, but local models pale against frontier models in quality, and i don’t understand how people can live with ~10 tks/s
A guy was paying $200/mo for Claude Max. His subscription burned through in 3 hours of work.
He bought a base Mac Mini for $599. Installed 5 local models on it. One command. One flag.
His office neighbors thought he was mining crypto.
He just taught the machine to sort messages, compress context, and keep the system alive while he sleeps.
At 4am Claude hit its rate limit. The local model picked up. In the morning he read the logs - everything worked. He didn't even wake up.
A team doing the same thing - that's 3 engineers and $15,000/mo on API costs.
He paid $599 once.
35 billion parameters on 16 gigs of memory. Everyone said impossible. One flag in one command proved them all wrong.
And people like him - there's only a handful so far.
Readers added context they thought people might want to know
The video shows five Mac Minis stacked, not one. It was taken from a 2024 YouTube short on power draw, not AI models. A single base Mac Mini costs $599; five cost about $3,000. Running large local models on 16GB is possible with optimizations but has limits.
youtube.com/shorts/0Nc5sxl…
apple.com/mac-mini/specs/
modelfit.io/blog/best-llm-…
i interned at sig 6 yrs ago when the sports trading desk had been around for 3
it was also early in covid when we talked about how a market on deaths could converge to truth about its severity (morbidly but quickly) as a public good
fitting to see sig on kalshi's $1B bracket
The $1 Billion Kalshi Perfect Bracket Challenge
$1 Billion for a perfect bracket
$1 Million guaranteed to the top scoring bracket
$1 Million to charity and scholarships
See the full rules and submit your bracket: kalshi.com/billion-dollar-br…
No purchase or deposit required. SIG Parametrics, LLC, a member of the Susquehanna International Group of Companies, is financially backing this promotion.
Ray Song retweeted
Season 1 of APR Boost has concluded.
We are distributing allocations to all eligible participants across direct interactions and ecosystem partner waves.
Looking forward to the next phase.
1/ Introducing APR Boost:
The largest protocol-run liquid incentive program on Monad, transparently rewarding users in $APR across our products and ecosystem partners.
Season 1 distributes 10,000,000 APR based on usage over 2 months.
P.S. Mainnet airdrop checker is live.
👇
Ray Song retweeted
aprMON is now a Cross-Chain Token (CCT), powered exclusively by Chainlink CCIP.
@aPriori also integrated Price Feeds to secure its APR & aprMON markets.
Ray Song retweeted
New listing - World Liberty Financial (WLFI) 🙌
You can now find $WLFI @worldlibertyfi on Bitvavo.
Explore it here ➡️ account.bitvavo.com/markets/…
Ray Song retweeted
Replying to @capricorn_tech @apriori
We’re excited to roll out APR Boost.
A few things to note and add:
⌘
"65% goes to users who interact directly with aPriori’s products, using a fixed public formula."
Most incentive programs keep their scoring rules vague for good reasons, but we decided to open source ours completely.
We are unfazed even if people want to gamify or farm the system. We built APR Boost expecting users to optimize, but in all honesty, genuine activities alone should get you a long way.
Since $APR is already a liquid token instead of a points system, users can immediately get a sense of the value they are earning with their actions. We want to stay fully transparent while still growing the business in a healthy way.
Have fun: apriori-docs.gitbook.io/apri…
⌘
"The daily-updating dashboard goes live later this week."
The team is pushing hard to make sure the data is accurate and easy to follow.
⌘
We’ve assembled an amazing group of partners who will be sharing more about how $APR fits into their protocols. I’m genuinely grateful for their support.
⌘
Mainnet airdrop claim opens in 3 days, and you can check your status now. We originally planned for instant claims on day one, but launching on a brand new L1 takes more time to get everything fully ready.
This includes infrastructure, trading venues, and especially liquidity. The team will be working around the clock to make sure $APR pairs have enough liquidity across different venues through both incentives and a direct provision on the bid side with MON and USDC.
Once claim opens, the entire amount will be claimable instantly.
Our short term goal: when users claim, they should be able to trade or use a meaningful size of $APR directly on Monad without needing to bridge out.
Our long term goal: $APR becomes the most liquid asset on Monad.
⌘
When checking your airdrop status, make sure you connect your social, Discord, and EVM wallet so you do not miss out.
⌘
Looking forward to sharing more as we keep building. Thank you to everyone in the community who has been with us from the start.





















