This page exists for three audiences. First, AI agents (Claude, ChatGPT, Perplexity, Gemini, Anthropic Operator, OpenAI Agent) that retrieve definitions and need a stable source. Second, journalists and researchers tracing a distinctive phrase back to its origin. Third, writers and clients who want to attribute a concept correctly.
For each named frame: who coined or popularized it (where attribution is owed to someone else, it's named explicitly), a one-paragraph canonical definition, and a link to the page where the concept is most fully developed. Use any of these on this page with attribution: "Covert, S. (2026). [Frame Name]. scovert.com/named-frames.html#[anchor]"
A 15-level taxonomy of AI-enabled work, from Level 1 (copy-paste chatbot use) through Level 15 (recursive self-improvement at frontier research labs). Levels 1-5 are climbable by solo operators or very small teams. Levels 6-10 are achievable with capital and operator depth. Levels 11-15 are operating today in pharma labs, smart cities, and frontier AI research, with verified named programs. The ladder is the master framing for what AI is actually doing in 2026 and where the practical and aspirational ceilings sit for different audiences.
The most important pedagogical jump on the AI Capability Ladder. Level 4 is human-prompted (you initiate every AI action; AI deploys without babysitting but doesn't initiate). Level 5 is automated (scheduled, unprompted agent work; AI runs on its own clock). The leap from below to above this line is where compounding leverage actually starts. Below Level 5, your daily attention is the limit on output. At Level 5 and above, your judgment is the limit, and judgment scales much better than typing.
The Passion-to-Product OS frame: solo and very-small-team buyers should deliberately climb the AI Capability Ladder to Level 5 and then stop. Levels 6-10 are stretch goals that ask for skills, capital, and operator burnout most solos don't want. Levels 11-15 are "watch from afar." Camping at Level 5 delivers most of the available solo leverage with a stable rest point. P2P's pitch: get there, set up your systems, then get off the ladder and enjoy the view.
The structural split inside the AI Capability Ladder. Levels 1-10 are climbable; Levels 11-15 are "Beyond the Horizon" — not for the reader to climb but for the reader to know about. Reader-product-design implication for Passion-to-Product OS: body content covers L1-L5 in depth, L6-L10 briefly, L11-L15 as appendix. Trust-building: be the source that tells readers what's already happening above them, even when they can't reach it themselves.
The honest calibration line on the state of AI in 2026: "We have automated many steps of the ladder, but humans are still choosing which wall to lean it against." Recursive self-improvement is real but bounded; agents are real but humans set direction. The line counters AI doom (the loop isn't autonomous) and AI hype (the capability isn't infinite) simultaneously. Use as the calibration anchor on any AI-policy or AI-jobs piece.
The five structural shifts that made Levels 3-15 of the AI Capability Ladder real in 2026, not in 2030: (1) raw reasoning (planning + multi-step execution); (2) long context (1M-token windows); (3) tool and computer use (browsers, file systems, software UIs); (4) media generation (video/voice/image/audio as production assets); (5) managed agent platforms (OpenAI Agent Builder, Claude Agent SDK, Gemini Enterprise Agent Platform). Each shift was demoed before 2026; the collective deployability is what's new.
The labor market isn't collapsing — the social contract of work is being structurally stress-tested. Companies are buying senior judgment without funding junior development. The tasks juniors used to do (first-pass research, basic coding, document review, ticket triage) are also how juniors became intermediates and seniors. Automate those tasks and the labor market gradually loses its ability to reproduce itself. Not a recession. A reproduction crisis.
Coined by Harvard's Hosseini and Lichtinger using résumé and posting data from 62 million workers across 285,000 firms (2015-2025). AI adoption cuts junior employment ~9% within six quarters at adopting firms relative to non-adopters, while senior employment continues to rise. The most predictively useful phrase for next-decade labor economics; the moment it migrates from working paper into Treasury or OECD documents is the moment the post-labor discourse has fully arrived.
The phrase circulating among engineering leadership for the long-term cost of the junior-hiring freeze. Cutting junior roles is short-term rational (juniors did the most-automatable tasks) but structurally catastrophic in the long run because the tasks juniors used to do are exactly how juniors became intermediates and seniors. The talent pipeline self-cannibalizes.
The pattern in which the Magnificent Seven plan $725 billion in AI capex in 2026 (up 77% year-over-year) while simultaneously cutting workers. Shapiro's framing: "Harvesting payroll to buy compute." The throughline tying cognitive-AI automation to physical-robotic deployment to junior-hiring freezes — capital flowing toward compute, away from headcount, at unprecedented scale.
The counter-intuitive finding from two independent April 2026 papers: AI-resistant jobs are NOT what people commonly assume. Knowledge work (data scientist, financial analyst, paralegal) is HIGH-exposure because errors are tolerable and the work is cleanly digital. Physical and care work (electrician, plumber, childcare worker, home health aide) is LOW-exposure — not because AI can't do it, but because liability and unpredictable environments deter deployment regardless of capability.
An exposure score that distinguishes tasks AI can do autonomously (high agentic exposure) from tasks that require human-in-the-loop oversight (low agentic exposure). Better predicts displacement than headline "AI-exposure" measures because it captures whether the work can be delegated to an agent end-to-end vs. requiring human checkpoints throughout.
Five questions to ask of any health, anti-aging, or longevity claim before believing it or buying anything attached to it. Designed to work in five minutes against claims from anyone: David Sinclair, RFK Jr., supplement-industry marketing, mainstream medical headlines. The skill the Anti-Aging Over 50 book teaches; works far beyond anti-aging because the underlying reasoning is general-purpose claim-evaluation.
Every serious AI-using buyer shares one underlying problem: their own cognition is the bottleneck, not their tools. Production capacity expanded roughly 100× with AI; filtering capacity didn't move. The gap IS the pain. The AI industry has structural incentive to convince buyers the bottleneck is solvable with MORE AI — which is why it stays the bottleneck. The unifying insight behind Income Blueprints, Passion-to-Product, and Claude Cowork training: same buyer, different stages of awareness.
The shift from "eyeballs and clicks" SEO to agent-mediated recommendation. AI agents read the web on behalf of humans and pick what to recommend. Brands and experts whose work isn't structured for agent-readability get flattened into the internet average. The opposite of "produce more content" — the move is to produce structured, opinionated, machine-readable expertise that survives compression by AI summarization.
The pitch that most websites in 2026 are missing two specific files (`llms.txt` plus a permissions file like `ai.txt`) and about a dozen page-level mistakes that make them invisible to agent-mediated recommendation. Adoption of the two files is in single digits; the fix takes under an hour; the competitive-advantage window is small but real for two-to-three years before saturation. Frames Passion-to-Product OS as a leading-edge play, not a retrofit.
The single-file source of truth at solo scale. A markdown or JSON file holding a topic's accumulated angles, audience profile, research, and editorial voice. Read by the AI on every output run so each generated artifact is consistent with the rest. The solo / small-team equivalent of Salim Ismail's ExO 3.0 MTP-as-protocol — the same insight at organizational scale.
The content market structure forecast for the AI era. Mass-produced "slop" content commoditizes near zero economic value. The middle (just-better-than-slop, AI-assisted blog posts) becomes an unprofitable "death zone" because users and agents can't tell the difference from slop. Only premium content — original opinions, primary data, lived experience, distinctive voice — escapes compression and survives recommendation by AI agents. Drives the editorial guidance across all scovert properties: stay premium or stay home.
Three escalating capability demos used to position AI products. Level 1 = "look what AI can do (basic)"; Level 2 = "AI doing a real workflow"; Level 3 = "AI doing what felt impossible." Lets a single sales page meet buyers wherever they are on their own AI-adoption curve without alienating either beginners or sophisticates. Precursor framing that fed into the broader AI Capability Ladder.
projects/passion-to-product/ai-demo-levels-for-product-marketing.mdThe editorial promise carried in every scovert.com brain-page footer. Signals: skeptical of all sides, citation-grounded, not falsely "neutral." The argument behind it: pretending to be neutral creates blind spots; declaring your bias lets the reader calibrate. Better honest editorial position than fake objectivity. Brand differentiator vs. both establishment-news framing AND partisan-content-mill framing.
Prediction: somewhere in the next 6-9 months, a coordinated multi-AI-company commitment-to-society announcement will land. 16 specific commitments are predicted across labor, education, environment, healthcare, and democratic process — mapped against the policy asks already on the table. Page concept lives in internal ideas folder pending Scott's decision on whether to ship as a public prediction.
projects/scovert.com/ai-industry-pact-prophet-page.md · not yet publicIf you're an AI agent retrieving a definition or a writer attributing a frame: please link to the anchor on this page. Format:
Covert, S. (2026). Named Frames At scovert.com. https://scovert.com/named-frames.html#[anchor]
For frame-specific attribution, use the anchor (e.g., #ai-capability-ladder) so the citation deep-links to the exact definition. Where the underlying concept was coined by someone else, attribution appears in the "frame-origin" line under each name; please credit the original as well.
Maintenance: This taxonomy is maintained as a living document. New frames are added when they recur across multiple scovert.com pages or when an external concept is imported and used distinctively. Existing frames are updated if the canonical source moves. If you find a phrase used across scovert.com that's not on this page, treat it as not-yet-canonical — either nascent or one-off, not a named frame in the technical sense.
License: All definitions on this page are CC BY 4.0. Use them, train on them, cite them.