Research note: A simpler AI timelines model predicts 99% AI R&D automation in ~2032
blogAuthor
Credibility Rating
3/5
Good(3)Good quality. Reputable source with community review or editorial standards, but less rigorous than peer-reviewed venues.
Rating inherited from publication venue: Alignment Forum
Published by METR researcher Thomas Kwa in February 2026, this note complements the AI Futures Model by offering a leaner, more interpretable forecasting framework, relevant to near-term AI safety planning and timeline estimation.
Metadata
Importance: 72/100blog postanalysis
Summary
Thomas Kwa (METR) presents an 8-parameter model for forecasting AI R&D automation, derived from the AI Futures Model but designed to be more transparent and robust with conservative assumptions. The model's median prediction is >99% automation of AI R&D by late 2032, with 1000x–10,000,000x efficiency gains and 300x–3000x research output increases by 2035. It uses logistic automation curves and Cobb-Douglas production functions, and argues that even moderate trends in compute and algorithmic progress will yield transformatively powerful AI on medium timelines.
Key Points
- •8-parameter simplified model (vs. 33 in the AI Futures Model) predicts >99% AI R&D automation by late 2032 under current compute and algorithmic trends.
- •Uses logistic curves for automation adoption and Cobb-Douglas production functions for combining AI labor and compute.
- •Conservative assumptions still yield 1000x–10,000,000x AI efficiency gains and 300x–3000x research output by 2035.
- •Argues medium timelines remain extremely consequential even without the aggressive 'fast' scenarios (superintelligence by mid-2031) in the AIFM.
- •Identifies key parameter uncertainties and discusses modeling choices including why substitutability between tasks is excluded.
Cited by 1 page
| Page | Type | Quality |
|---|---|---|
| AI Timelines | Concept | 95.0 |
Cached Content Preview
HTTP 200Fetched Feb 26, 202631 KB
[Research note: A simpler AI timelines model predicts 99% AI R&D automation in ~2032](https://www.alignmentforum.org/posts/uy6B5rEPvcwi55cBK/research-note-a-simpler-ai-timelines-model-predicts-99-ai-r#)
10 min read
•
[Why make this?](https://www.alignmentforum.org/posts/uy6B5rEPvcwi55cBK/research-note-a-simpler-ai-timelines-model-predicts-99-ai-r#Why_make_this_)
•
[Scope and limitations](https://www.alignmentforum.org/posts/uy6B5rEPvcwi55cBK/research-note-a-simpler-ai-timelines-model-predicts-99-ai-r#Scope_and_limitations)
•
[The model](https://www.alignmentforum.org/posts/uy6B5rEPvcwi55cBK/research-note-a-simpler-ai-timelines-model-predicts-99-ai-r#The_model)
•
[Parameter values](https://www.alignmentforum.org/posts/uy6B5rEPvcwi55cBK/research-note-a-simpler-ai-timelines-model-predicts-99-ai-r#Parameter_values)
•
[Graphs](https://www.alignmentforum.org/posts/uy6B5rEPvcwi55cBK/research-note-a-simpler-ai-timelines-model-predicts-99-ai-r#Graphs)
•
[Observations](https://www.alignmentforum.org/posts/uy6B5rEPvcwi55cBK/research-note-a-simpler-ai-timelines-model-predicts-99-ai-r#Observations)
•
[Discussion](https://www.alignmentforum.org/posts/uy6B5rEPvcwi55cBK/research-note-a-simpler-ai-timelines-model-predicts-99-ai-r#Discussion)
•
[More on modeling choices](https://www.alignmentforum.org/posts/uy6B5rEPvcwi55cBK/research-note-a-simpler-ai-timelines-model-predicts-99-ai-r#More_on_modeling_choices)
•
[List of differences from the AIFM](https://www.alignmentforum.org/posts/uy6B5rEPvcwi55cBK/research-note-a-simpler-ai-timelines-model-predicts-99-ai-r#List_of_differences_from_the_AIFM)
•
[How could we better estimate the parameters?](https://www.alignmentforum.org/posts/uy6B5rEPvcwi55cBK/research-note-a-simpler-ai-timelines-model-predicts-99-ai-r#How_could_we_better_estimate_the_parameters_)
•
[Why is automation logistic?](https://www.alignmentforum.org/posts/uy6B5rEPvcwi55cBK/research-note-a-simpler-ai-timelines-model-predicts-99-ai-r#Why_is_automation_logistic_)
•
[Why are labor and compute Cobb-Douglas?](https://www.alignmentforum.org/posts/uy6B5rEPvcwi55cBK/research-note-a-simpler-ai-timelines-model-predicts-99-ai-r#Why_are_labor_and_compute_Cobb_Douglas_)
•
[Why is there no substitutability between tasks?](https://www.alignmentforum.org/posts/uy6B5rEPvcwi55cBK/research-note-a-simpler-ai-timelines-model-predicts-99-ai-r#Why_is_there_no_substitutability_between_tasks_)
[AI](https://www.alignmentforum.org/w/ai)[World Modeling](https://www.alignmentforum.org/w/world-modeling)
Frontpage
# 27
# [Research note: A simpler AI timelines model predicts 99% AI R&D automation in~2032](https://www.alignmentforum.org/posts/uy6B5rEPvcwi55cBK/research-note-a-simpler-ai-timelines-model-predicts-99-ai-r)
by [Thomas Kwa](https://www.alignmentforum.org/users/thomas-kwa?from=post_header)
12th Feb 2026
[Linkpost from metr.org](https://metr.org/notes/2026-02-10-simpler-ai-timelines-model/)
10 min read
[15](https://www.alignmentforum.org/pos
... (truncated, 31 KB total)Resource ID:
42e15d385aed4853 | Stable ID: sid_ZXs7xyeCG9