The AI Task Frontier

Horizon, forgiveness, and directed AI innovation. The paper studies why the profitable direction of AI progress changes across task regions.

June 2026 Theory paper 99 math checks passed 19-page fuller economics draft
Open PDF LaTeX source Math checker Learning log

Main Intuition

AI progress is not one scalar quality ladder. Tasks fail for different reasons, projects move different task boundaries, and cheap AI products change the incentives of frontier labs. The current draft connects the task-level project-value rule to a two-period frontier race with distillers and low-cost followers.

Task Frontier

B_H values horizon progress, B_F values reliability in unforgiving environments, and project value is scarcity times movement per dollar.

Project Inversion

The same project menu can favor runtime, tools, or search in one task region and reliability or verification in another. The regional ratio B_H^R/B_F^R determines the direction.

Frontier Race

Cheap followers erode saturated-market rents, which can push foundational labs toward longer, higher-value, harder-to-copy task regions.

Revision Plan & Today's Learning

Current Plan

  • Keep the paper framed as a full economics theory draft, not a theorem catalog.
  • Use the dynamic game only to explain frontier-lab incentives and distillation pressure.
  • Keep the main text centered on four economic propositions: project value, cheap-product sorting, frontier racing, and distillation direction.
  • Keep mechanical derivations in the appendix.

Today's Learning: June 28-30

  • Rounds 54-91 clarified the core object: task-side project value, not a generic AI competition story.
  • Round 91 rewrote the abstract around the economic puzzle, the horizon-by-forgiveness primitive, and the project-value rule.
  • Round 92 compressed the paper from a technical compendium into a 13-page readable theory draft.
  • Round 92 removed competition, data-flywheel, stopping, private-social, and infrastructure extensions from the main PDF.
  • Round 93 added readability scaffolding: a cleaner abstract, clearer scope, intuition before formulas, and examples before the main theorem.
  • Round 94 expanded the paper into a fuller economics draft with a two-period frontier race and distillation/cheap-player channel.
  • The appendix now contains short proofs and mechanical derivations.
Open full learning log from today

This panel is synced from the project-level learning.md log for the June 28-30 revision sequence.

Task Regions

High forgiveness
Low forgiveness
Low horizon
Product recommendation, ad copy, search help. Low-cost systems and imitation can be strong.
Driving maneuvers, medical dosage, payments control. Reliability and verification dominate.
High horizon
Coding, debugging, data analysis, research assistance. Runtime, tools, retries, and workflow integration matter most.
Drug discovery, critical infrastructure, autonomous organizations. Full-stack frontier advantages can persist.

Paper

The PDF preview is unavailable in this browser. Open the paper directly.

The compiled PDF, LaTeX source, and math checker in this folder are synchronized with the current revision.