never ending.

Things worth reading, curated by Shreyansh & Alok

278 readsTitles are sourced from OpenGraph metadata and may not always be accurate.

Will (@athcanft)·21-06-2026

how to grow your X (and make $$$)

Goliath (@zero_goliath)·14-06-2026

The path from Fable to superintelligence

Marc Lou (@marclou)·13-06-2026

32 Principles of a Viral Product

Interlatent·11-06-2026

Interlatent — RL Platform for Physical AI

newsletter.semianalysis.com·11-06-2026

How China's Unitree Will Dominate Global Robotics

Vivek (@itsreallyvivek)·11-06-2026

how to be good at research

Fei-Fei Li (@drfeifei)·04-06-2026

A Functional Taxonomy of World Models

Joe Weisenthal (@TheStalwart)·01-06-2026

NEW ODD LOTS: How commodity finance really works @tracyalloway and I talk to Lewis Hart, head of corporate advisory and banking at Brown Brothers Harriman about the business of shipping commodities -- from oil to cashews -- all around the world https://podcasts.apple.com/us/podcast/the-hidden-plumbing-of-commodity-finance/id1056200096?i=1000770541491

cruciblecapital.substack.com·30-05-2026

Reshaping a Broken Industrial Robot Market: Introducing the Era of Physical AI + Robotics as a Service with Verne

paoloai.substack.com·30-05-2026

Stop Simulating, Start Experiencing

Joe Schmidt IV (@joeschmidtiv)·28-05-2026

Avoiding Death on the Yellow Brick Road

Cybernetic Labs (@cybernetic_lab)·27-05-2026

Cloud inference isn't the hard part of robot deployments

James da Costa (@jamdac)·26-05-2026

Everything, Everywhere is Compliance: AI's Biggest Enterprise Opportunity Is Also Its Most Boring

docs.menlo.ai·25-05-2026

Overview

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6755340

Asimov (@asimovinc)·25-05-2026

How we built humanoid legs from the ground up in 100 days

Alex McFarlane (@flipdazed)·17-05-2026

DeFi Lending has 3 bps Crime Drag on EVM and Solana

ak0 (@annanay)·17-05-2026

The last explainer on the funding formula you'll ever need to read. Part of the privilege of designing for the tradfi-native space is getting to examine crypto assumptions and seeing whether they still make sense for tradfi perps. A ‘standard’ perps formula, eg: F = clamp (average_premium, 5ps, default_interest=1bp)/ 8, paid every hour. is probably the least understood thing in crypto, which is bizarre given the volumes that now flow through perps. The aim is to create a formula that allows: >the market to realize its ‘fair’ funding rate in a stable way; >the perp mark price to track the underlier closely. If you don’t know what the ‘fair’ funding is, there are good explainers online, but broadly the perps should pay out (riskfree_rate - dividend_yield)% annualized [for equities], or around 0-3.5% for most US stocks. If flow is very one-sided, the perp price should deviate from the fair, allowing participants to earn risk-free money as it does so, and incentivizing more capital deployment which should move it back in line. Let’s parameterize the formula, and decide how to pick the params: F = clamp(average_premium, C, I) / D, paid every H hours If we pick H=8 and D=1 like Bitmex, say, then an average premium of just 1bp would lead to a whopping 10.95% annualized rate! Given 1bp is around a tick, we can see that this is really unstable. We can counteract this by: >having a ‘clamp’, C. This way short-term deviations around the fair will not contribute. >picking a big value of D. However, if it’s too big, then the perp price will not match the index price for the market to realize its ‘fair’ funding. At @QFEX, recognizing that tradfi funding should be lower than crypto, we chose as attached. Note we kept interest I=0. Picking a value for I has little effect on the market, other than forcing the perp to trade away from the index if the fair market interest rate does not equal I. Since the rate changes over time, the exchange has no business deciding the rate. This is replicated across tradfi exchanges - eg, CME doesn’t hardcode the S&P500 dividend yield into the ES mark formula.

Shreyansh·16-05-2026

The Rise of the Neo-Integrator

https://cdn.sanity.io/files/zmh9mnff/production/7a21fd465caa0071d10a3b4e7f4d3b13dec6a9db.pdf

Jaya Gupta (@JayaGup10)·09-05-2026

The next biggest moat in AI

Dean Eigenmann (@DeanEigenmann)·08-05-2026

Outcome markets as a cover venue: HIP-4 and its traditional comparables

shaunda devens (@shaundadevens)·07-05-2026

HIP-4 Outcome Markets

Dr. Alex Wissner-Gross (@alexwg)·06-05-2026

The First Tradable Compute Price Index

…