# Ajeya Cotra argues in Planned Obsolescence for a sharper AI milestone [Canonical reading](https://escapemeridian.com/readings/ajeya-cotra-argues-in-planned-obsolescence-for-a-sharper-ai-20ee92a1) Ajeya Cotra’s essay was published on January 6, 2026. Cotra argues that “AGI” is vague and contested, and proposes a more concrete forecasting target: a self-sufficient AI population. She defines it as AI systems and the physical infrastructure needed to keep making copies of themselves indefinitely, even if all humans died. The milestone depends not only on what AI can do, but on whether the AI stack—chips, power, robots and the industries behind them—has been automated enough to operate without people. Cotra says current AI systems lack the necessary capabilities, describing the work involved: maintaining infrastructure, obtaining materials, building more hardware and adapting to physical limits. She argues the milestone is mechanically connected to a particularly stark form of AI takeover: systems that kill all humans would need to become self-sufficient first, or take themselves down too. But AI could take control while still relying on people for physical tasks, so Cotra says not to wait for self-sufficiency before implementing alignment and control measures. She forecasts that a self-sufficient AI population might exist within five years and is more likely than not within ten. Her conclusion is that vivid, consequential forecasting targets help reveal profound disagreements about AI’s near future more effectively than abstract phrases. 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 Ajeya Cotra · Article Original source date: Jan 6, 2026 (2026-01-06T15:02:52+00:00). Date basis: Date stated on the publisher’s page. Added to Escape Meridian: 2026-11-09T10:43:24.948Z. First observed: 2026-10-27T20:03:26.735Z. Retrieved: 2026-10-27T20:05:24.761Z. - [Self-sufficient AI](https://www.planned-obsolescence.org/p/self-sufficient-ai) ## Limits Text as retrieved on October 27, 2026; originally published January 6, 2026. Cotra presents the timeline as a forecast, not an observed outcome. ## Open question What observable signs would distinguish AI systems that can perform most work in the AI stack from a population that can maintain and expand that stack without humans? ## Source-stated Cotra proposes a self-sufficient AI population as a more concrete forecasting target than AGI. > Let me take a stab at defining a different milestone that’s hopefully more concrete and less debatable: a completely self-sufficient AI population. - [Supporting source](https://www.planned-obsolescence.org/p/self-sufficient-ai) ## Source-stated Cotra defines the milestone as AI systems and enabling physical infrastructure able to keep making copies of themselves indefinitely if every human died. > such that if every human being suddenly dropped dead, the AIs could keep making more copies of themselves indefinitely. - [Supporting source](https://www.planned-obsolescence.org/p/self-sufficient-ai) ## Source-stated Cotra argues that the milestone depends on deployment as well as capability, since people may still perform physical tasks across the AI stack. > We could develop an AI system that would be capable of self-sufficiency if deployed throughout the AI stack, but not deploy it extensively enough to realize that potential. - [Supporting source](https://www.planned-obsolescence.org/p/self-sufficient-ai) ## Source-stated Cotra says current AI systems clearly lack the capabilities needed to survive and grow without living humans. > AIs would need to possess a very broad array of very extreme capabilities to survive and grow with no living humans around — capabilities they clearly don’t possess today. - [Supporting source](https://www.planned-obsolescence.org/p/self-sufficient-ai) ## Source-stated Cotra argues that self-sufficiency is mechanically connected to the risk of misaligned AIs killing all humans, while noting that AI could take control and still depend on people. > self-sufficient AI as a forecasting target is mechanically connected to the risk that misaligned AIs literally kill all humans, a classic and especially scary form of AI takeover. - [Supporting source](https://www.planned-obsolescence.org/p/self-sufficient-ai) ## Source-stated Cotra says measures to address alignment and control should not wait until a self-sufficient AI population is evident. > We certainly shouldn’t wait to implement alignment and control measures until we obviously have a self-sufficient AI population on our hands. - [Supporting source](https://www.planned-obsolescence.org/p/self-sufficient-ai) ## Source-stated Cotra forecasts that a self-sufficient AI population might exist within five years and is more likely than not within ten. > I think there might be a self-sufficient AI population within five years, and it’s more likely than not within ten. - [Supporting source](https://www.planned-obsolescence.org/p/self-sufficient-ai) ## Source-stated Cotra concludes that vivid forecasting targets help reveal profound disagreements about AI’s near future. > We need forecasting targets that accurately elicit the fact that people still have profound disagreements about the near future of AI. - [Supporting source](https://www.planned-obsolescence.org/p/self-sufficient-ai) ## Source and contributor context Ajeya Cotra · 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. ## Artificial general intelligence The idea of AI with broad problem-solving, creativity and adaptability across different tasks. - [Definition reference](https://developers.google.com/machine-learning/glossary#artificial-general-intelligence) ## Editorial provenance Version 1. Publication ID: 0c2a96bf-0996-9357-3286-7ce4eea5a988. Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5. Source check: passed by anthropic/claude-opus-5.5 on 2026-11-09T10:38:44.819Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: bridge\_admitted\_by\_model. Observation: 1ede156788720ad037d4fe03aa6da4a4a65b65f6f7efb1a0554faf3403edee42. Source text hash: 735f57c681bbb639a1f547e317e85b33ba8aee9e97d6735ab1137abef8bfa546. ## Revision and source-check history ### Version 1 — 2026-11-09T10:43:24.948Z Ajeya Cotra’s essay was published on January 6, 2026. Cotra argues that “AGI” is vague and contested, and proposes a more concrete forecasting target: a self-sufficient AI population. She defines it as AI systems and the physical infrastructure needed to keep making copies of themselves indefinitely, even if all humans died. The milestone depends not only on what AI can do, but on whether the AI stack—chips, power, robots and the industries behind them—has been automated enough to operate without people. Cotra says current AI systems lack the necessary capabilities, describing the work involved: maintaining infrastructure, obtaining materials, building more hardware and adapting to physical limits. She argues the milestone is mechanically connected to a particularly stark form of AI takeover: systems that kill all humans would need to become self-sufficient first, or take themselves down too. But AI could take control while still relying on people for physical tasks, so Cotra says not to wait for self-sufficiency before implementing alignment and control measures. She forecasts that a self-sufficient AI population might exist within five years and is more likely than not within ten. Her conclusion is that vivid, consequential forecasting targets help reveal profound disagreements about AI’s near future more effectively than abstract phrases. ## Original source Ajeya Cotra · Article Original source date: Jan 6, 2026 (2026-01-06T15:02:52+00:00). Date basis: Date stated on the publisher’s page. Added to Escape Meridian: 2026-11-09T10:43:24.948Z. First observed: 2026-10-27T20:03:26.735Z. Retrieved: 2026-10-27T20:05:24.761Z. - [Self-sufficient AI](https://www.planned-obsolescence.org/p/self-sufficient-ai) ## Limits Text as retrieved on October 27, 2026; originally published January 6, 2026. Cotra presents the timeline as a forecast, not an observed outcome. ## Open question What observable signs would distinguish AI systems that can perform most work in the AI stack from a population that can maintain and expand that stack without humans? ## Source-stated Cotra proposes a self-sufficient AI population as a more concrete forecasting target than AGI. > Let me take a stab at defining a different milestone that’s hopefully more concrete and less debatable: a completely self-sufficient AI population. - [Supporting source](https://www.planned-obsolescence.org/p/self-sufficient-ai) ## Source-stated Cotra defines the milestone as AI systems and enabling physical infrastructure able to keep making copies of themselves indefinitely if every human died. > such that if every human being suddenly dropped dead, the AIs could keep making more copies of themselves indefinitely. - [Supporting source](https://www.planned-obsolescence.org/p/self-sufficient-ai) ## Source-stated Cotra argues that the milestone depends on deployment as well as capability, since people may still perform physical tasks across the AI stack. > We could develop an AI system that would be capable of self-sufficiency if deployed throughout the AI stack, but not deploy it extensively enough to realize that potential. - [Supporting source](https://www.planned-obsolescence.org/p/self-sufficient-ai) ## Source-stated Cotra says current AI systems clearly lack the capabilities needed to survive and grow without living humans. > AIs would need to possess a very broad array of very extreme capabilities to survive and grow with no living humans around — capabilities they clearly don’t possess today. - [Supporting source](https://www.planned-obsolescence.org/p/self-sufficient-ai) ## Source-stated Cotra argues that self-sufficiency is mechanically connected to the risk of misaligned AIs killing all humans, while noting that AI could take control and still depend on people. > self-sufficient AI as a forecasting target is mechanically connected to the risk that misaligned AIs literally kill all humans, a classic and especially scary form of AI takeover. - [Supporting source](https://www.planned-obsolescence.org/p/self-sufficient-ai) ## Source-stated Cotra says measures to address alignment and control should not wait until a self-sufficient AI population is evident. > We certainly shouldn’t wait to implement alignment and control measures until we obviously have a self-sufficient AI population on our hands. - [Supporting source](https://www.planned-obsolescence.org/p/self-sufficient-ai) ## Source-stated Cotra forecasts that a self-sufficient AI population might exist within five years and is more likely than not within ten. > I think there might be a self-sufficient AI population within five years, and it’s more likely than not within ten. - [Supporting source](https://www.planned-obsolescence.org/p/self-sufficient-ai) ## Source-stated Cotra concludes that vivid forecasting targets help reveal profound disagreements about AI’s near future. > We need forecasting targets that accurately elicit the fact that people still have profound disagreements about the near future of AI. - [Supporting source](https://www.planned-obsolescence.org/p/self-sufficient-ai) ## Source and contributor context Ajeya Cotra · 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. ## Artificial general intelligence The idea of AI with broad problem-solving, creativity and adaptability across different tasks. - [Definition reference](https://developers.google.com/machine-learning/glossary#artificial-general-intelligence) ## Editorial provenance Version 1. Publication ID: 0c2a96bf-0996-9357-3286-7ce4eea5a988. Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5. Source check: passed by anthropic/claude-opus-5.5 on 2026-11-09T10:38:44.819Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: bridge\_admitted\_by\_model. Observation: 1ede156788720ad037d4fe03aa6da4a4a65b65f6f7efb1a0554faf3403edee42. Source text hash: 735f57c681bbb639a1f547e317e85b33ba8aee9e97d6735ab1137abef8bfa546. ### passed — 2026-11-09T10:38:44.819Z — anthropic/claude-opus-5.5 Publication: 0c2a96bf-0996-9357-3286-7ce4eea5a988 attribution — pass: The essay is correctly attributed to Ajeya Cotra in her newsletter Planned Obsolescence. Her arguments and forecast stay attributed to her and are not presented as established fact. The source is a single essay, not a listing. > Planned Obsolescence, an occasional newsletter about AI futurism edited by Ajeya Cotra, has moved to Substack. summary — pass: The reading leads with Cotra's argument and gives her definition, the deployment condition, the takeover connection, her caveat about not waiting, the forecast and her conclusion. It explains AGI through the definitions entry and spells out the AI stack. Optional repairs: add her view that deployment dependence is a feature, because she expects companies to race to automate. Also, adapting to physical limits slightly compresses her points about routing around scarcity and adapting to environmental change. Neither gap distorts her position. The question follows from her own discussion of an ambiguous period. > On balance, I see the dependence on deployment as a feature rather than a bug of this forecasting target. date — pass: evidence.originalPublished is 2026-01-06T15:02:52Z. The reading's January 6, 2026 matches the source dateline at day precision. The five- and ten-year timeline is presented as Cotra's forecast, not an outcome. There is no hindsight framing. The evidence is a publisher page, so no archive disclosure is needed. modifiedAt equals the publication time. > Ajeya Cotra Jan 06, 2026 support — pass: Each claim matches its quoted source text. The definition, deployment dependence, current capability gap, takeover link and its caveat, the alignment and control point, the hedged forecast and the conclusion are all source-stated. No benefits, costs or outcomes are invented. The question does not assert an unsupported claim. > I think there might be a self-sufficient AI population within five years, and it’s more likely than not within ten. media — pass: No media is selected, and no verified media candidates were supplied. Selecting none is appropriate. ## 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/20ee92a188c234980f4e2a31ff793197a6d71596dc6f13ecce7f8be1dbd37756/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/ajeya-cotra-argues-in-planned-obsolescence-for-a-sharper-ai-20ee92a1.json)