According to the AI Futures Project "AI 2027", AI coding will surpass any living person's coding abilities in January 2027. This will resolve as Yes on Feb 1 2027 if there are no demonstrable economically valuable coding tasks that a person can do that an AI cannot do. It will resolve as No if there are any economically valuable tasks that a human can do that AI cannot do.
Evidence of an Economically Valuable Coding Task Will Include:
- Any legitimate benchmark that is not saturated by AI, and contains any task a human can complete independently (such as SWEBench).
- Any repeatable demonstration of such task
Source:
https://ai-2027.com/ → Scroll to January 2027 and hover on the Coding bar, which says AI will have a score of 2.1, indicating it has surpasses all living people.
Update 2025-04-12 (PST) (AI summary of creator comment): Scope of Role Replacement:
The resolution is now clarified to compare AI agents against professional developers in the context of full software engineering roles, not just individual competitive coding tasks.
Full Replacement Requirement: The market will resolve as Yes if an AI Agent can replace any human developer 100% in all their technical work (coding, designing architecture, monitoring systems, etc.).
Exemption: AI research–specific technical work is exempt from this requirement.
High Bar: This clarification establishes a high bar consistent with the AI 2027 document's claim that the AI agent will be better than any living human developer.
1,000
3.00Being superhuman at coding is more than just performing well at individual tasks. It requires executive function for task prioritization etc.
This is likely to resolve YES without it actually leading to a software intelligence explosion
@Siebe In the AI 2027 document, it is comparing AI agents to professional developers in the context of software development, not competitive coding or anything like that. So this is super human at the job of software engineer.
So this resolves Yes if an AI Agent can replace any human developer 100% in all their technical work. The only technical work exempt is AI research specific work (as that’s a distinction in the document).
This means the agent needs to be able to code, design architecture, monitors systems, priorice work, debug prod, etc.
I’m setting a high bar because the document explicitly states the agent will be better than anyone alive.
@HawkeyePierce okay! I think that's the right call to make, but different from what your resolution criteria implied (to me at least)
What is this?
What is Manifold?
Are our predictions accurate?
How do I win cash prizes?
and can't be cashed out. to participate and winners can withdraw sweepcash as a cash prize. You can filter for sweepstakes markets on the browse page.1.00Related questions
What is this?
What is Manifold?
Are our predictions accurate?
How do I win cash prizes?
and can't be cashed out. to participate and winners can withdraw sweepcash as a cash prize. You can filter for sweepstakes markets on the browse page.1.00