Capably Human launches framework for building ‘Capability Fitness’ with AI

Capably Human launched in Windsor, Ontario, with a free research library aimed at making AI use strengthen, rather than replace, human capability. The company is opening its framework, model and engagement tools to researchers, educators, employers and product teams as AI adoption rises and concerns grow about skill erosion.

Why it matters: - Capably Human is trying to define a new standard for human-AI work that protects judgment, learning and independence. - The company’s framework frames AI as a design choice that can either build capability or replace the effort that creates it. - The launch lands as more workers use AI at work and more people worry about losing critical thinking and other skills.

What happened: - Capably Human launched today in Windsor, Ontario, with its research library open to the public. - The company’s free framework, The Capability Fitness Framework (CH-F-001), defines Capability Fitness and distinguishes AI help that builds capability from help that replaces it. - The companion model, The Capability Model (CH-M-001), maps the 26 skills most affected by working with AI. - Both are available free at capablyhuman.com/publications.

The details: - Capably Human researches, designs and tests how people work with AI so both become more capable. - The company defines Capability Fitness as the lifelong development and maintenance of the mind’s capabilities to reason and decide, learn, focus, create and adapt. - The framework is built on five propositions: capability develops through practice, technology changes patterns of practice, Capability Fitness is multidimensional, development is lifelong and contextual, and measurement should support development rather than rank people. - The framework separates support, where the person keeps the meaningful work, from substitution, where AI removes the effort that builds capability. - It sets eight Human-Strengthening AI design principles, including Attempt Before Assistance, Questions Before Answers and Capability Over Dependency. - The model organizes capability into five domains: Think, Learn, Focus, Create and Adapt. - The model breaks those domains into 26 sub-skills, including Problem Framing, Evidence Evaluation, Persistence Through Difficulty and Original Voice. - Each sub-skill is measured on four levels: Prompted, Guided, Independent and Transferred. - Transfer, meaning a skill that shows up beyond where it was practiced, is the level that counts. - Each sub-skill includes an AI failure mode that shows how delegating too early can replace practice that builds capability. - The framework says some skills matter more because of AI, including Question Formation, since AI output depends on the quality of the question. - Capably Human’s rules of engagement cover three decisions in any human-AI interaction: delegate or retain, trust or challenge, and let AI run or step in. - The company is also releasing The Rules of Engagement Check (CH-I-001), The Rules of Engagement Model (CH-M-002) and The Rules of Engagement Guide (CH-G-002), with examples for work, school and later life. - The company compares its intended AI role to a spotter in the gym: helping people do more, but not lifting the bar for them.

Between the lines: - Capably Human is pushing back on the dominant pattern of AI use, which often optimizes for speed and finished output over long-term learning. - The company cites Gallup data showing 53% of U.S. employees use AI at work. - It also cites a 2026 IBM Institute for Business Value survey in which 60% of employees feared their skills eroding, especially critical thinking. - Early randomized studies cited by the company suggest that the way AI help is designed changes outcomes. - In one example, Turkish high schoolers using a standard AI chatbot did 48% better on practice problems but 17% worse on a later exam without it. - In that example, a version that gave hints rather than answers performed about as well as practicing without AI. - In three online experiments described in a 2026 preprint, ten minutes of AI help left people solving 57% of later problems alone, compared with 73% without AI. - In both examples, the losses concentrated where AI gave answers. - The company says its publications are provisional and states what evidence would count against the framework. - The model’s main gap is not yet filled: observable anchors for each proficiency level still need to be written and tested.

What’s next: - Capably Human plans trials comparing AI drafting first versus the person attempting first, explaining versus asking questions, agreeing versus challenging, and full delegation versus structured hand-offs. - The company says the trials will measure performance, learning, judgment and independence. - Trial designs will be published first, then run with consenting alpha members of the company’s first product. - Results will appear as Capably Human Reports. - A book is planned for December 2026. - The waitlist for the Humably alpha is open now, with the alpha scheduled to start in Q1 2027. - Researchers, educators, product teams, employers and policymakers are invited to use, test and critique the library.

The bottom line: - Capably Human is trying to turn AI from a shortcut into a training ground for human capability.