# Brad DeLong argues in The Attention Info-Bio Tech Economy that Claude is a tool, not a person [Canonical reading](https://escapemeridian.com/readings/brad-delong-argues-in-the-attention-info-bio-tech-economy-that-e0fc4486) Brad DeLong’s piece was published on October 13, 2026. He agrees with John Gruber and Ben Thompson that Anthropic’s restriction on repeated, purposeless cruelty toward Claude treats an AI tool unlike an ordinary tool. In the crossposted discussion, Thompson compares cursing at AI to cursing at a tool after an accident; Gruber imagines being locked out of a spreadsheet for swearing at Excel. DeLong calls Claude and similar systems “stochastic-parrot clever-hans software tools,” and argues that their ability to produce human-like text does not make them conscious or a “Digital God.” He says people can overread that ability, while the discussion also considers why language makes AI seem human. DeLong’s own conclusion is that AI can be useful and potentially destructive as a human social-cognitive information technology without being a godlike being. He allows, however, that “stochastic parrot” may no longer capture what properly trained and harnessed LLMs do when woven into human cognition and public reason. How much of human thought this change involves remains an open question for DeLong. Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims. ## Original source J. Bradford DeLong · Article Original source date: Oct 13, 2026 (2026-10-13T19:41:43.000Z). Date basis: Date supplied by the publisher’s feed. Added to Escape Meridian: 2026-11-04T23:03:23.617Z. First observed: 2026-10-24T08:19:55.880Z. Retrieved: 2026-10-24T08:21:37.768Z. - [CROSSPOST: JOHN GRUBER & BEN THOMPSON: Meanness & Claude](https://braddelong.substack.com/p/crosspost-john-gruber-and-ben-thompson) ## Limits Text as retrieved on October 24, 2026; originally published October 13, 2026. The discussion about Claude’s restriction is presented through DeLong’s crosspost and the quoted conversation; the supplied evidence does not include the linked episode itself. ## Open question What line have properly trained and harnessed LLMs crossed as they enter human cognition and public reason, and how much of human thought does that involve? ## Source-stated Brad DeLong says he agrees with John Gruber and Ben Thompson’s criticism of Anthropic’s restriction on extreme, repeated cruelty toward Claude. > I am on Ben and John’s side here: These are stochastic-parrot clever-hans software tools - [Supporting source](https://braddelong.substack.com/p/crosspost-john-gruber-and-ben-thompson) ## Source-stated DeLong argues that people may mistake AI’s ability to produce human-like text for consciousness or godlike status. > my response whenever these things exceed my expectations with respect to their ability to do things that humans do makes me increase the amount of human stuff I judge to be stochastic parrotage, rather than to raise my subjective probability that Anthropic is actually on the road to building Digital - [Supporting source](https://braddelong.substack.com/p/crosspost-john-gruber-and-ben-thompson) ## Source-stated DeLong concludes that AI may be useful and potentially destructive as a human social-cognitive information technology without giving rise to a Digital God. > But that is because of it usefulness and potential destructiveness as a human social-cognitive information technology. It is not because it will give rise to a Digital God. - [Supporting source](https://braddelong.substack.com/p/crosspost-john-gruber-and-ben-thompson) ## Source-stated DeLong says the “stochastic parrot” metaphor may no longer describe what properly trained and harnessed LLMs do in human cognition and public reason; he identifies the extent of that change as an open question. > Properly trained, RAGed, and harnessed LLMs may still be stochastic parrot, but as they are enmeshed in the spheres of human cognition and public reason they increasingly do not act as stochastic parrot. They have crossed some sort of line - [Supporting source](https://braddelong.substack.com/p/crosspost-john-gruber-and-ben-thompson) ## Source-stated DeLong’s piece is dated October 13, 2026. > Brad DeLong > Oct 13, 2026 - [Supporting source](https://braddelong.substack.com/p/crosspost-john-gruber-and-ben-thompson) ## Source and contributor context J. Bradford DeLong · Publisher The publisher is identified here. No individual author has been inferred. ## Source media Media retrieval status: unchecked. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Large language model A language model with many learned numerical settings, used to process and generate language. - [Definition reference](https://developers.google.com/machine-learning/glossary#large-language-model) ## Claude Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding. - [Explore this context](/entities/claude) - [Reference](https://claude.com/product/overview) - [Reference](https://www.anthropic.com/news) - [Reference](https://www.anthropic.com/research/team/economics) ## Anthropic Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems. - [Explore this context](/entities/anthropic) - [Reference](https://www.anthropic.com/news) - [Reference](https://www.anthropic.com/research/team/economics) ## Editorial provenance Version 1. Publication ID: d68e7f32-e055-d995-852d-bc70703a687b. Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-04T22:58:25.140Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: bridge\_admitted\_by\_model, second\_opinion\_categorical. Observation: e90203919fec16a169fb25cfc3abb041e272d193eb7a6e024acc720bc2d17c34. Source text hash: 29367f1ddba458c6fec32069f22980a432fd8aaf1ee5ae0a29a39aaa86044394. ## Revision and source-check history ### Version 1 — 2026-11-04T23:03:23.617Z Brad DeLong’s piece was published on October 13, 2026. He agrees with John Gruber and Ben Thompson that Anthropic’s restriction on repeated, purposeless cruelty toward Claude treats an AI tool unlike an ordinary tool. In the crossposted discussion, Thompson compares cursing at AI to cursing at a tool after an accident; Gruber imagines being locked out of a spreadsheet for swearing at Excel. DeLong calls Claude and similar systems “stochastic-parrot clever-hans software tools,” and argues that their ability to produce human-like text does not make them conscious or a “Digital God.” He says people can overread that ability, while the discussion also considers why language makes AI seem human. DeLong’s own conclusion is that AI can be useful and potentially destructive as a human social-cognitive information technology without being a godlike being. He allows, however, that “stochastic parrot” may no longer capture what properly trained and harnessed LLMs do when woven into human cognition and public reason. How much of human thought this change involves remains an open question for DeLong. ## Original source J. Bradford DeLong · Article Original source date: Oct 13, 2026 (2026-10-13T19:41:43.000Z). Date basis: Date supplied by the publisher’s feed. Added to Escape Meridian: 2026-11-04T23:03:23.617Z. First observed: 2026-10-24T08:19:55.880Z. Retrieved: 2026-10-24T08:21:37.768Z. - [CROSSPOST: JOHN GRUBER & BEN THOMPSON: Meanness & Claude](https://braddelong.substack.com/p/crosspost-john-gruber-and-ben-thompson) ## Limits Text as retrieved on October 24, 2026; originally published October 13, 2026. The discussion about Claude’s restriction is presented through DeLong’s crosspost and the quoted conversation; the supplied evidence does not include the linked episode itself. ## Open question What line have properly trained and harnessed LLMs crossed as they enter human cognition and public reason, and how much of human thought does that involve? ## Source-stated Brad DeLong says he agrees with John Gruber and Ben Thompson’s criticism of Anthropic’s restriction on extreme, repeated cruelty toward Claude. > I am on Ben and John’s side here: These are stochastic-parrot clever-hans software tools - [Supporting source](https://braddelong.substack.com/p/crosspost-john-gruber-and-ben-thompson) ## Source-stated DeLong argues that people may mistake AI’s ability to produce human-like text for consciousness or godlike status. > my response whenever these things exceed my expectations with respect to their ability to do things that humans do makes me increase the amount of human stuff I judge to be stochastic parrotage, rather than to raise my subjective probability that Anthropic is actually on the road to building Digital - [Supporting source](https://braddelong.substack.com/p/crosspost-john-gruber-and-ben-thompson) ## Source-stated DeLong concludes that AI may be useful and potentially destructive as a human social-cognitive information technology without giving rise to a Digital God. > But that is because of it usefulness and potential destructiveness as a human social-cognitive information technology. It is not because it will give rise to a Digital God. - [Supporting source](https://braddelong.substack.com/p/crosspost-john-gruber-and-ben-thompson) ## Source-stated DeLong says the “stochastic parrot” metaphor may no longer describe what properly trained and harnessed LLMs do in human cognition and public reason; he identifies the extent of that change as an open question. > Properly trained, RAGed, and harnessed LLMs may still be stochastic parrot, but as they are enmeshed in the spheres of human cognition and public reason they increasingly do not act as stochastic parrot. They have crossed some sort of line - [Supporting source](https://braddelong.substack.com/p/crosspost-john-gruber-and-ben-thompson) ## Source-stated DeLong’s piece is dated October 13, 2026. > Brad DeLong > Oct 13, 2026 - [Supporting source](https://braddelong.substack.com/p/crosspost-john-gruber-and-ben-thompson) ## Source and contributor context J. Bradford DeLong · Publisher The publisher is identified here. No individual author has been inferred. ## Source media Media retrieval status: unchecked. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## Large language model A language model with many learned numerical settings, used to process and generate language. - [Definition reference](https://developers.google.com/machine-learning/glossary#large-language-model) ## Claude Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding. - [Explore this context](/entities/claude) - [Reference](https://claude.com/product/overview) - [Reference](https://www.anthropic.com/news) - [Reference](https://www.anthropic.com/research/team/economics) ## Anthropic Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems. - [Explore this context](/entities/anthropic) - [Reference](https://www.anthropic.com/news) - [Reference](https://www.anthropic.com/research/team/economics) ## Editorial provenance Version 1. Publication ID: d68e7f32-e055-d995-852d-bc70703a687b. Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-04T22:58:25.140Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: bridge\_admitted\_by\_model, second\_opinion\_categorical. Observation: e90203919fec16a169fb25cfc3abb041e272d193eb7a6e024acc720bc2d17c34. Source text hash: 29367f1ddba458c6fec32069f22980a432fd8aaf1ee5ae0a29a39aaa86044394. ### uncertain — 2026-11-04T22:58:25.140Z — anthropic/claude-opus-5.5 Publication: d68e7f32-e055-d995-852d-bc70703a687b attribution — pass: The reading attributes the framing argument to Brad DeLong and the tool and Excel comparisons to Thompson and Gruber within the crossposted transcript. DeLong's stance is his own. Naming the header 'The Attention Info-Bio Tech Economy' as the venue is drawn from the page text and does not misattribute any claim. > I am on Ben and John’s side here: These are stochastic-parrot clever-hans software tools summary — uncertain: Central terms are not explained for a general reader: 'stochastic parrot,' 'clever-hans' and 'Digital God' are neither glossed nor in the definitions; only LLM is defined. Suggest adding term-library entries. The question also asks 'What line' LLMs have crossed, but DeLong answers that himself. His open issue is in what sense 'enough' of the human minds behind the training data is summoned, not how much of human thought is involved. Smallest repair: reframe the question around that 'enough' issue and gloss the key terms. > That of summoning from the Vasty Deep enough of the human minds whose actions formed their training data. In what sense “enough”? Ah. That is the big question now. date — pass: originalPublished is 2026-10-13T19:41:43Z, which matches the dateline Oct 13, 2026 stated at day precision. modifiedAt equals the publication time. The reading does not rely on any later passages and uses no hindsight framing. The evidence is a publisher page, not an archive copy. > Brad DeLong Oct 13, 2026 support — pass: Each claim matches the source: DeLong sides with Gruber and Thompson; he treats impressive outputs as evidence of more stochastic parrotage, not of a Digital God; he says AI matters as a social-cognitive technology, not as a Digital God; and he amends the stochastic parrot metaphor. The reading adds no unsupported benefits or costs. > I still cannot see how any other judgment is possible except for people who have let the term “Artificial Intelligence” hack their brains, to their confusion. media — pass: No media was selected and none was supplied or verified, so there is no identity or relevance issue. ## Recorded references Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter. No public references recorded here yet. [Reference data (JSON)](/readings/e0fc44869edc5535193777bf5c379d6abf3a892d4b26d18e0602a60552f4ba06/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/brad-delong-argues-in-the-attention-info-bio-tech-economy-that-e0fc4486.json)