What will be the METR AI coding uplift multiple in 2026?

MANIFOLD

Ṁ10kṀ37k

2027

<0.8x

0.8-1x

1-1.2x

17%

1.2-1.4x

25%

1.4-1.6x

18%

1.6-1.8x

11%

1.8-2x

2-2.5x

2.5-3x

≥3x

This market matches Developer Productivity: METR Uplift Study from the AI 2026 Forecasting Survey by AI Digest.

See other manifold questions here

Resolution criteria

Resolves to the speedup multiple (time_taken_by_developers_without_AI / time_taken_by_developers_with_AI) reported by the most recent METR uplift study published in 2026 using the same or clearly comparable methodology. If METR reports percent change in time taken instead (time_taken_by_developers_with_AI / time_taken_by_developers_without_AI - 1) as they did for their 2025 study, we will convert using: speedup = 1 / (1 + percent_change_in_time_taken).

If no METR uplift study is published in 2026, this question's outcome will be marked ambiguous. If multiple comparable METR uplift studies are published in 2026, the resolver may decide which one is most representative of the end of 2026 software development speedup, at their discretion (this would be the one published later in 2026, unless there was reason to believe it was less comparable to the 2025 study than an earlier one).

Eli Lifland is responsible for final judgment on resolution decisions.

Buckets are left-inclusive: e.g., 1.2-1.4x includes 1.2x but not 1.4x.

Market context

Technology

2026 AI Forecasting Survey by AI Digest

Get

1,000

to start trading!

People are also trading

February 2026 AI releases

What will happen in 2026 related to AI?

Will the METR AI coding uplift market be resolved based on the Feb 2026 updated estimate?

18% chance

Best AI time horizon by February 2026, per METR?

Best AI time horizon by August 2026, per METR?

By what percentage will using AI slowdown/speedup developers in the second METR study?

What will be the METR time horizon doubling time in 2026?

Will AI (large language models) collapse by may 2026?

9% chance

As predicted by AI 2027, will AI surpass all living humans in coding ability by January 2027?

11% chance

Will AI progress surprise Metaculus?

Sort by:

update:

sold Ṁ250 NO

@Bayesian meta market please on whether this market resolves according to this study or a subsequent one

@Bayesian When I filled out my forecast2026.ai for some reason I just thought it resolved to the results of the next METR uplift study 💀

@jim oops no it's to the latest one of 2026 i think, i agree that was kind of a bad phrasing on my end
anyway market is here:

opened a Ṁ200 YES at 8% order

@Bayesian >3x seems underpriced then 😃

opened a Ṁ5,000 NO at 10% order

@jim limit up

@Bayesian If this market we’re getting resolved today what would the correct answer be? Where did the number 0.84 come from?

@MaxLennartson 0.84 came from the METR uplift study from early 2025 where experienced open source devs were slowed down by using AI tools working in their existing large projects.

@DavidHiggs I know that their was a 20% decrease but how did they get from 20% to 0.84?

@MaxLennartson A 20% decrease = x - 0.2x = 0.8x

@dreev How do I read the new METR report in terms of uplift?

18% decrease in completion time means someone that took 100 minutes without AI would take 82 minutes with AI right? so that's a 100/82 = 21.9% speedup, or 1.219x speedup? If i'm not goofing

previously they slowed down by 19%, so someone who took 100 minutes without AI took 119 minutes with AI. so that's a 100/119 speedup, or a 0.84x speedup. so yeah i'm pretty sure the above is correct and the current measured AI coding speedup for experienced developers is 1.22x . but the new update has the experienced OSS coders population and some other more general population iiuc, i can't speak to which one would count for the purpose of this market if this study had been the last one of 2026