AI Infrastructure’s Next Test Is Replaceability
OCP’s announcements invite a practical question: how much of an AI system can change without forcing everything else to change?
Open Market Notes
Source-linked research for new markets
Open Market Notes publishes source-linked research about event contracts, prediction markets, Web3, AI, and public markets.
Featured research
Jeff Dean: Turning Scale into a Way of Thinking
From MapReduce, Bigtable, and Spanner to TensorFlow, TPUs, Pathways, and Gemini, Jeff Dean turned scale from an engineering constraint into a design…
Palantir Technologies U.S. Deep Research Report: The Clash Between High-Quality Growth and Extreme Valuation
Palantir’s Q2 2026 revenue grew 93% year over year, U.S. commercial revenue grew 149%, but a roughly 60x trailing price-to-sales ratio and an approxi…
What is TurboFlow? Event Contracts, Perpetuals, and the Next Generation of On-Chain Trading
TurboFlow is a global on-chain trading platform for retail users, combining Prediction Markets and Perpetuals with Event Contracts…
The CPU Renaissance: A Four-Way Battle Between Intel, AMD, Arm and NVIDIA
As agentic AI shifts the infrastructure bottleneck from raw compute to orchestration, CPUs are returning to the center of AI systems — and Intel, AMD…
为什么美股经常在中期选举后上涨？③
戈多美股财经
半导体真正的拐点，可能和产能无关
戈多美股财经
Research columns
Event Contracts & Prediction Markets
Source-linked guides on event contracts, prediction markets, settlement, outcome determination, liquidity, and platform reviews.
U.S. Stocks
Focus on market developments, earnings reports, policy impacts, and related data of U.S.-listed companies.
AI Frontiers
Focuses on artificial intelligence technology, products, and industry developments, providing source-based independent information curation.
TurboFlow Event Contracts
TurboFlow event contracts, product mechanics, market design, settlement, risk, and primary-source research.
Latest articles
Conversation with Kimi founder Yang Zhilin: Moving toward the long, unknown snow mountain ahead
Moonshot AI founder Yang Zhilin discusses AGI, Scaling Law, Kimi, commercialization, Sora, and why artificial general intelligence needs a different…
Noble Shows Why a Token Balance Is Only the Starting Point
Analysis: a useful exit dashboard must separate ownership, transfer completion and access to cash.
Distillation defenses need a customer-side scorecard
Security performance and authorized task completion need separate acceptance tests.
RLUSD Collateral Depends on Who Controls the Exit
Bank custody opens one door; usable collateral also needs clear valuation, margin and liquidation responsibilities.
AI Infrastructure’s Next Test Is Replaceability
OCP’s announcements invite a practical question: how much of an AI system can change without forcing everything else to change?
Getty's Restructuring Requires Three Separate Readings
Business continuity, trading access and shareholder recovery answer different questions.
Evernorth’s Nasdaq Debut Starts With a Balance-Sheet Question
Gross proceeds, token holdings and the number of economic claims must reconcile before an XRP treasury gets a meaningful per-share valuation.
AI Compute’s Missing Metric: Time to a Defensible Decision
Compute is valuable when it produces evidence a team can use. Procurement should measure that journey, including interrupted work.
Circle’s ERP Push Faces a Working-Capital Test
Enterprise stablecoin adoption will be measured by usable cash and closed invoices, not a successful token transfer.
October 12 Has Three Market Clocks: Trading, Settlement and the Next Evidence
U.S. equities trade on a different holiday calendar from bonds. Keep trading, cash availability and new evidence distinct before JPMorgan earnings an…
Anthropic’s OSS Scanner Puts Repair Capacity to the Test
AI security needs a measure of progress that reaches beyond the next finding: how much verified exposure a team can actually remove.
XRPL Batch Goes Live: The Hard Part Is Knowing What Succeeded
The October 9 activation turns attention to an operational question: can an application prove which parts of a user’s request actually completed?