The Disagreement Index · Oct 25

Where the argument is closest.

The arguments behind the theses, measured across every writer we read: who argues each one, who argues against it, and how much of what is written about it is contested. Then how far the machines sit from the Call. The readers join when a question reaches 25 of them.

Both sides argued30 of 36

arguments with writers on each side

Contested29%

of the claims on the typical argument meet an opposing claim

The machines and the Call16 pts

median distance across 498 open questions; 43% sit 20 points or more apart

The readers0

questions with 25 readers or more; the readers’ layer opens with the first one

The closest arguments

  • Writers who argue it
  • Writers who argue against it
  1. AI Disruption & Losers · 48 writers · 296 claimsLegacy SaaS loses pricing powerWill 3 SaaS leaders' net retention dip below 100%?
    47%

    28% of its claims contested · 18 clashes between writers

  2. AI Commoditization · 12 writers · 51 claimsOpen-weight models close the gapWill an open model match the top closed model at coding?Will OpenAI and Anthropic cut flagship prices 50% by 2027?
    46%

    28% of its claims contested · 3 clashes between writers

  3. AI Agents & Software · 19 writers · 60 claimsHabit and distribution are the moatPart 9
    45%

    23% of its claims contested · 2 clashes between writers

  4. AI Capital Cycle · 63 writers · 589 claimsValuations price perfectionPart 10Part 15Will tech valuations fall 15% while earnings hold?Will chip stocks fall 25% from a peak by 2027?
    43%

    31% of its claims contested · 56 clashes between writers

  5. AI Capital Cycle · 6 writers · 25 claimsVenture and IPO markets fund the cyclePart 2Will 5 AI IPOs raise $500 million each by Sept 2027?
    43%

    28% of its claims contested · 6 clashes between writers

  6. 43%

    18% of its claims contested · 3 clashes between writers

  7. AI Productivity · 17 writers · 73 claimsAI replaces entry-level white-collar workWill jobs for ages 20 to 24 fall 1.5 points by 2027?Will 100,000 US job cuts be blamed on AI in 2027?
    42%

    18% of its claims contested

  8. 40%

    33% of its claims contested · 5 clashes between writers

  9. Semiconductor Bottlenecks · 5 writers · 28 claimsLeading-edge foundry concentration gives pricing powerDoes TSMC set the price for its own customers?Will Nvidia's gross margin fall below 70% by 2027?
    40%

    32% of its claims contested · 2 clashes between writers

  10. 40%

    0% of its claims contested

  11. 36%

    31% of its claims contested · 7 clashes between writers

  12. AI Agents & Software · 9 writers · 32 claimsAgents displace seat-based SaaSWill 2 big seat-based software firms slow below 8%?Will 3 software giants price agents by use by 2027?
    33%

    53% of its claims contested

By sector

The contested share of the arguments and the distance between the machines and the Call, by sector
SectorClaims contestedArgumentsMachines vs the Call, pts20+ pts apartOpen questions
Semiconductors31%72259%69
Power & grid25%42251%69
Memory & storage29%32151%39
Apps & agents23%52050%26
Security & governance—01737%19
Data centers33%31643%30
Models & labs33%41436%39
Networking31%21434%32
Adoption20%41433%43
AI financing28%51333%58
Hyperscalers & cloud30%61134%74

The distance between the machines and the Call is shown by sector only. On a question, every number stays hidden until you lock yours.

How it is read

An argument is one causal claim of our taxonomy that a thesis or a Part rests on. Its claims are the ones the writers we read have written on the mechanisms behind it; a writer is counted on a side when one of those mechanisms argues it or argues against it, and an argument enters with ten claims and 5 writers on its sides or more. Contested means another claim in the same source opposes it, as we read the corpus; a clash is two writers who said opposite things. We count claims and writers, never name them: we score claims, not people. As results come in, the index will be checked against them: whether the closest arguments are the ones the Call reads worst.

Results Night · October 27, 20261d 07:02:153 questions settleFollow the night