Ben Thompson · Jan 13, 2025
Ben Thompson argues in Stratechery that AI’s arrival will be uneven
Markdown · JSON · Original source
Ben Thompson’s article was published January 13, 2025. He argues that AI systems able to take and complete tasks could change how work gets done, but that the change will reach companies unevenly. He points to o3’s performance on the ARC-AGI test as a sign of progress in inference- scaling. He contrasts these systems with traditional , which he describes as tools for people who bring their own expertise and judgment. Thompson’s central argument is that many established companies may struggle to use because their work depends on experience and tacit knowledge that documented processes do not capture. He compares SaaS companies’ per-seat licensing to advertising that measures a proxy for the outcome rather than the outcome itself. He expects new companies built around AI to be better positioned to use agents as “ammunition.” By contrast, he says that AI’s broad adoption by established companies with real-world differentiation could take years. His conclusion is that the future may arrive unevenly: larger, successful companies may benefit less in the short term, while companies that start without established workforces may be better placed to use AI. How much of established companies’ work can AI agents complete when critical knowledge remains tacit?
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
Ben Thompson · Article
Source details
Original source date: Jan 13, 2025 (2025-01-13T14:59:09+00:00).
Date basis: Date stated on the publisher’s page.
Added to Escape Meridian: 2026-11-16T00:58:24.424Z. First observed: 2026-11-14T07:57:47.835Z. Retrieved: 2026-11-14T07:59:38.734Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
Text as retrieved on November 14, 2026; originally published January 13, 2025. The article’s claims about o3’s task performance and future AI adoption are presented as Thompson’s arguments and expectations; this evidence does not establish how perform across company work.
Open question
How much of established companies’ work can complete when critical knowledge remains tacit?
Source-stated
Ben Thompson argues that o3’s performance on ARC-AGI reflects progress in inference-time scaling, with performance varying according to the compute allocated.
o3 significantly outperforms o1, and the extent of that outperformance is dictated by how much computing is allocated to the problem at hand.
Source-stated
Thompson argues that traditional LLMs are best treated as tools for people who can apply expertise and judgment to manage mistakes.
the biggest returns come to those with the most expertise and agency, who are able to use their own knowledge and judgment to reap efficiency gains while managing hallucinations and mistakes.
Source-stated
Thompson argues that established companies may struggle to adopt AI agents because their processes depend on tacit human knowledge.
No matter how well-documented a company’s processes might be, it will become clear that there are massive gaps that were filled through experience and tacit knowledge by the human ammunition.
Source-stated
Thompson argues that new companies may be more able to use AI agents, while broad adoption by established companies could take years.
the true AI takeover of enterprises that retain real world differentiation will likely take years.
Source-stated
Thompson concludes that AI’s benefits may arrive unevenly, with larger, successful companies potentially benefiting less in the short term.
the larger and more successful a company is the less they may benefit in the short term.
Source-stated
The source dates the article to January 13, 2025.
Monday, January 13, 2025
Source and contributor context
Ben Thompson · 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.
AI agent
An AI system that chooses and carries out actions toward a goal, often using software tools.
TIME
TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Editorial provenance
Version 1. Publication ID: a4400a99-39f2-8c75-4a54-bb1e9757a2b7.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-16T00:55:35.048Z. 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: 1a394156e5e4d9babf10d8ea255f15ff02fd3481b39bc6604134671e71793505. Source text hash: 469c6477a6b9e89361cacbd48cad27e4679c9484642a1f2ae5fd9302fc31a229.
Revision and source-check history
Version 1 · 2026-11-16T00:58:24.424Z
Ben Thompson argues in Stratechery that AI’s arrival will be uneven
Ben Thompson’s article was published January 13, 2025. He argues that AI systems able to take and complete tasks could change how work gets done, but that the change will reach companies unevenly. He points to o3’s performance on the ARC-AGI test as a sign of progress in inference- scaling. He contrasts these systems with traditional , which he describes as tools for people who bring their own expertise and judgment. Thompson’s central argument is that many established companies may struggle to use because their work depends on experience and tacit knowledge that documented processes do not capture. He compares SaaS companies’ per-seat licensing to advertising that measures a proxy for the outcome rather than the outcome itself. He expects new companies built around AI to be better positioned to use agents as “ammunition.” By contrast, he says that AI’s broad adoption by established companies with real-world differentiation could take years. His conclusion is that the future may arrive unevenly: larger, successful companies may benefit less in the short term, while companies that start without established workforces may be better placed to use AI. How much of established companies’ work can AI agents complete when critical knowledge remains tacit?
Original source
Ben Thompson · Article
Source details
Original source date: Jan 13, 2025 (2025-01-13T14:59:09+00:00).
Date basis: Date stated on the publisher’s page.
Added to Escape Meridian: 2026-11-16T00:58:24.424Z. First observed: 2026-11-14T07:57:47.835Z. Retrieved: 2026-11-14T07:59:38.734Z.
Limits
Text as retrieved on November 14, 2026; originally published January 13, 2025. The article’s claims about o3’s task performance and future AI adoption are presented as Thompson’s arguments and expectations; this evidence does not establish how perform across company work.
Open question
How much of established companies’ work can complete when critical knowledge remains tacit?
Source-stated
Ben Thompson argues that o3’s performance on ARC-AGI reflects progress in inference-time scaling, with performance varying according to the compute allocated.
o3 significantly outperforms o1, and the extent of that outperformance is dictated by how much computing is allocated to the problem at hand.
Source-stated
Thompson argues that traditional LLMs are best treated as tools for people who can apply expertise and judgment to manage mistakes.
the biggest returns come to those with the most expertise and agency, who are able to use their own knowledge and judgment to reap efficiency gains while managing hallucinations and mistakes.
Source-stated
Thompson argues that established companies may struggle to adopt AI agents because their processes depend on tacit human knowledge.
No matter how well-documented a company’s processes might be, it will become clear that there are massive gaps that were filled through experience and tacit knowledge by the human ammunition.
Source-stated
Thompson argues that new companies may be more able to use AI agents, while broad adoption by established companies could take years.
the true AI takeover of enterprises that retain real world differentiation will likely take years.
Source-stated
Thompson concludes that AI’s benefits may arrive unevenly, with larger, successful companies potentially benefiting less in the short term.
the larger and more successful a company is the less they may benefit in the short term.
Source-stated
The source dates the article to January 13, 2025.
Monday, January 13, 2025
Source and contributor context
Ben Thompson · 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.
AI agent
An AI system that chooses and carries out actions toward a goal, often using software tools.
TIME
TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Editorial provenance
Version 1. Publication ID: a4400a99-39f2-8c75-4a54-bb1e9757a2b7.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-16T00:55:35.048Z. 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: 1a394156e5e4d9babf10d8ea255f15ff02fd3481b39bc6604134671e71793505. Source text hash: 469c6477a6b9e89361cacbd48cad27e4679c9484642a1f2ae5fd9302fc31a229.
uncertain · 2026-11-16T00:55:35.048Z · anthropic/claude-opus-5.5
Publication: a4400a99-39f2-8c75-4a54-bb1e9757a2b7
attribution — pass: The reading consistently attributes the argument to Ben Thompson writing in Stratechery, which matches the publisher and the first-person essay. Quoted material from ARC Prize, Sam Altman and Keith Rabois is not misattributed to Thompson.
The reason for the extended digression on advertising, however, is to explain why I’m skeptical about AI “materially chang[ing] the output of companies”, at least in 2025.
summary — uncertain: Several central terms go unexplained for a general reader, and the page definitions cover only LLM and AI agent. The reading never says o3 is an OpenAI model, never explains inference-time scaling (more compute while answering improves accuracy) or ARC-AGI (a puzzle test easy for humans and hard for LLMs), and never defines SaaS. It quotes "ammunition" without Rabois's barrels-and-ammunition metaphor, in which ammunition means a worker who can be handed a task and trusted to finish it. Smallest repair: add one plain clause for each, or add them to the term library.
The promise of AI generally, and inference-time scaling models in particular, is that they can be ammunition
date — pass: originalPublished is 2025-01-13T14:59:09Z, which matches the dateline Monday, January 13, 2025, and the reading states the date at day precision. The evidence kind is publisher-page, so no archive disclosure is needed. The trailing links to related later pieces are not used. Forecasts such as adoption taking years are framed as Thompson's expectations, and there is no hindsight framing.
Monday, January 13, 2025
support — pass: Each claim matches the source's text and keeps it hedged as Thompson's argument. This covers o3's compute-dependent outperformance, traditional LLMs as tools for experts, the tacit-knowledge gaps, years-long enterprise adoption, and smaller short-term benefit for large firms. The line about companies that start without established workforces reflects his closing line. The limitations note correctly says the evidence does not establish agent performance.
the entities that will truly leverage AI, however, will not be the ones that replace them, but start without them.
media — pass: No media is selected, and no verified candidates were supplied, so there is no media identity or relevance issue.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Explore TIMEA language model with many learned numerical settings, used to process and generate language.
Definition referenceAn AI system that chooses and carries out actions toward a goal, often using software tools.
Definition referenceAn AI system that chooses and carries out actions toward a goal, often using software tools.
Definition referenceAn AI system that chooses and carries out actions toward a goal, often using software tools.
Definition referenceTIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Explore TIMEA language model with many learned numerical settings, used to process and generate language.
Definition referenceAn AI system that chooses and carries out actions toward a goal, often using software tools.
Definition referenceAn AI system that chooses and carries out actions toward a goal, often using software tools.
Definition referenceAn AI system that chooses and carries out actions toward a goal, often using software tools.
Definition reference