SecondSourceJudgment rebuilt from primary sources

How we checked the items below

For readers who want to dig in. What happened, in brief, is on the Today tab.

This issue draws on SecondSource's own research brief and last night's collection. Part of last night's collection is still being rerun, so the source counts in this issue are provisional.

1. [Today] (index, October 25) On Ornn's transaction index, B200 and H200 rental prices rose more than 20% in a month and passed a provider's list price, while H100 lagged

Not proven yet: transaction prices come from one index provider, Ornn

How we checked

The index readings come from Ornn's home page. The home page doesn't say which providers the sample covers or how large each deal is, and Ornn's methodology page hasn't been reviewed. Daily readings from Silicon Data, another index provider, aren't available for now. In a September interview, Ornn CTO Wayne Nelms said the index tracks only part of the market. He also said larger rentals command higher unit prices, because very few sellers can deliver ten thousand or more interconnected GPUs at once (The Cognitive Revolution interview). If the index includes deals of that size, transaction prices above list could reflect deal size rather than tighter supply. The available data can't separate the two. Intercontinental Exchange has announced plans to list compute futures based on this index, pending regulatory approval; this issue carries no link to that announcement. Rising compute rental prices benefit Ornn, the company that compiles the index.

Back to this item ↑

[This week] (post, October 23) Gregory C. Allen, director of the AI center at the US think tank CSIS, argues that complete export controls on AI chips to China would not cost NVIDIA a cent of revenue.

How we checked

The post doesn't say which company's earnings call it was.

(Gregory C. Allen, 2026-10-23)

Back to this item ↑

[Looking back] (SecondSource judgment card, September 6) Did OpenAI publicly explain its German-website incident by October 18? On the material available, no.

How we checked

(judgment card)

Back to this item ↑

Still missing today. The gap that would most affect judgment is Ornn's methodology page: which providers the sample covers and how large each deal is haven't been reviewed. Daily H100 readings from Silicon Data, another index provider, aren't available for now, so the transaction prices have no second reading to compare against. Other gaps: the full text of the three preprints, the 253 company filings and the 4 macroeconomic releases haven't been reviewed.

Sources & accounting

The past 24 hours. 802 new pieces came in from October 25 to 26: 253 company filings, 224 academic papers, 141 company and personal blog posts, 93 social-platform posts, 56 from a second academic-paper feed (counted separately from the academic papers above), 16 show transcripts, 15 subscription newsletters and 4 macroeconomic data releases. Of those 802, 184 were read in full, 120 were screened out as off-topic, 496 haven't been read yet, and 2 are waiting to be fetched again. Of the academic papers, 71 were read, 31 screened out and 122 not yet read. Of the blog posts, 44 were read, 30 screened out and 67 not yet read. Of the social-platform posts, 41 were read, 50 screened out and 2 are waiting to be fetched again. Of the show transcripts, 15 were read and 1 screened out. None of the 253 company filings or the 4 macroeconomic releases has been read. From the second paper feed, 7 were read; of the newsletters, 6 were read. Last night's data collection stopped partway through and was rerun early this morning, so these figures are readings taken before the rerun finished. Most of the new blog entries are older posts from July to September listed again; company blogs had few new posts today. Last night's collection picked up 334 original X posts; those 334 are counted differently from the 93 above, so the two can't be divided. One account, @bhorowitz, accounts for 49 of the 334, followed by @teortaxesTex, @elonmusk, @Miles_Brundage and @GaryMarcus. Shows included The Cognitive Revolution, Latent Space, ChinaTalk, Training Data and Dwarkesh Podcast; newsletters included Latent Space, AI Supremacy, Ahead of AI, Exponential View, Gary Marcus and Interconnects.

One-time catch-up. No new one-time catch-up of older material today.

Tracked sources. The long-term roster currently has 529 named sources: 302 on social platforms, 90 shows, 51 news outlets, 48 blogs, 48 paper authors and 46 newsletters, with the rest spread across earnings reports, keynotes and other channels. The same source may be tracked on several channels, so the channel figures add up to more than 529. Those are counts of tracked sources and can't be added to or compared with the 802 new pieces. Representative names: on social platforms, Lucas Beyer, Graham Neubig, Daniel Kokotajlo and Arvind Narayanan; among newsletters, Zvi Mowshowitz, Miles Brundage and Dean Ball; on the shows roster, interviewees such as Sam Altman, Dario Amodei and Demis Hassabis. This issue uses 7 outside sources in the body. Only outside links actually cited in the body are counted, and links to SecondSource's own site are left out; this is a different population from the 529 tracked sources above. On the Taiwanese side, 7 new sources were added in September, bringing the total to 135; Taiwanese media, Taiwanese companies' official channels and local reporters in Taiwan count, while foreign outlets relaying Taiwanese media do not.