Glossary · 42 ideas
The named ideas and frameworks I use across my writing on AI, work, markets and technology. Each has a plain definition and links to the articles where it comes from.
The path a standalone AI feature travels from novel product to native checkbox: someone wraps a workflow in a model call, the demand is proven, then the model provider or platform ships the same capability natively for free.
A scaffold is AI that supports your thinking, surfaces information and handles low-value processing, freeing attention for deeper work. A surrogate replaces your own reasoning, analysis and judgment and gradually hollows out the cognitive capacity you are trying to protect.
Also: the three-stage arc
The shape of how AI shows up inside a company: the assistant answers when asked (you use it), the agent reasons and triggers the next step (you work alongside it) and the coworker owns a slice of work end to end (you manage it). The article labels this arc informed extrapolation.
Also: the Brochure Phase
Using AI the way companies used the internet in 1995: brochure-level moves (chatbots on websites, automated meeting notes, dashboards labelled AI-powered) that are incremental and safe, instead of rebuilding the process with AI as the foundation.
Why today's digital documents fail machines: they are visual, not semantic; meaning is implicit, not explicit; they are unstructured or only pretend to be structured; they can't do anything by themselves; and there is no single source of truth.
The combination of Projects, structured MD files and task-specific Agents that turns a capable AI into a consistent collaborator. It is not a product; it is infrastructure, and it separates vibe coding that scales from vibe coding that frustrates.
Three eras of documents: the Paper Age (1950 to 2000, typewriters and carbon copies), the Pixel Age (2000 to 2025, Word, PDF and Google Docs built for human eyes) and the Action Age (2025 to 2030, executable objects designed for robots first, humans second).
A business whose correction cycle runs at the speed of its AI harness loops rather than the speed of its meeting calendar: the loop senses a miss, drafts the fix, checks it against governance rules and ships it or hands a one-line decision to a human.
The failure mode that occurs when fabricated training data, created with legitimate scientific intent, becomes indistinguishable from observed data in downstream model training pipelines, because provenance metadata is stripped and lineage is lost.
Also: token wall
The point deep in a build where you hit the token limit, the context window closes and the momentum breaks, like a paywall after the first paragraph. Called the most underreported real cost of vibe coding.
Also: Human Layer and Machine Layer
A split world for documents: a Human Layer of presentation views (PDFs, web pages, Word docs) on top of a Machine Layer where the real document lives as structured data, executable clauses, cryptographic signatures and live API bindings.
Every major business function runs its own AI team that knows its process deeply (the whales), and all of those teams roll into one central AI organization (the pod) that owns shared standards, governance, security and reusable components.
The belief that because you can build the foundation yourself, you should. Security, identity, integrations, content management, workflow automation and customer data infrastructure are the floor, not your competitive advantage.
Network effects remain the strongest defensibility available, but in AI SaaS the participation layer that builds them now has a running meter, so margin becomes the constraint on how fast you can afford to buy the network.
A simple self-disruption model: launch the new product aggressively only if it beats the old one by 1.5x after costs. The extra 0.5x covers the Risk Tax (new bets can flop and cost money), Friction Drag (some customers won't switch) and Decay (the old product is already going stale).
Also: two-engine operating model
A way for incumbent SaaS companies to report AI: engine one is the core system of record (stable, profitable, funds the business); engine two is the AI expansion layer (usage-based, adoption-focused, prioritized for growth over margin).
To find an interest that matters, commit to 90 days of consistent practice at four to six hours a week: long enough to test, short enough to pivot.
The era in which purpose replaces paychecks. Interests are sustained engagements that require curiosity, judgment, craft and relational investment; they are what we do when survival is no longer negotiated month to month by wage labor.
If the Singularity is a threshold of speed, the modern job, as the primary source of identity and livelihood, structurally collapses, bringing an economic shock and a psychological shock at the same time.
Organizations redesign their go-to-market in months and their compensation plans in years, because comp is where an organization's real beliefs live. You have a Comp Lag when your pricing and your sales comp plan sit on different rungs of the seats, consumption, outcomes ladder.
The boundary between work whose demand expands when its cost falls and work whose demand is capped no matter how cheap it gets. It decides which side of the AI tool you are on.
The observation that the wage-earning employee, working for a single large organization on a stable salary, is a roughly 150-year artifact of the industrial era rather than the historical default of human work.
AI isn't just fast, it's exposing: it reveals how much of our day was filled with tasks that didn't require us, just our presence and our compliance, and gives back the space to bring humanity back into our work.
Founders building sophisticated companies with significant revenue and thousands of customers without a single employee, using AI, automation and no-code tools to do the work of entire departments.
Individuals curating and validating their own skills, knowledge and achievements through tangible, verifiable proof of work, built on a proof-of-work portfolio instead of relying on universities as gatekeepers.
The space between what your job was filling and what you actually need to feel directed. Work provides income, but also structure, identity and something that functions like purpose.
The gap between the value a severance agreement could deliver and the value you actually capture, lost not because you lacked leverage but because you treated a negotiable contract as a verdict and signed on the first read.
Divest from the market you can no longer compete in, then use national security language, not trade language, to make sure the wall around that market stays high and gets harder to remove.
Also: money, supply and expectations
Strip away the jargon and inflation comes down to three dials moving at once: the money dial (how fast money grows relative to the economy), the supply dial (whether the world can deliver what people want) and the expectations dial (whether people believe prices will keep rising and price that in).
A two-to-five-year window where AI eliminates income faster than housing prices can adjust and far faster than governments can respond. It is three problems moving at different speeds: AI displacement (fast), housing repricing (slow) and policy response (slowest).
Also: Venture Investment in Singular Tech Agents
Venture Investment in Singular Tech Agents: backing AI-first visionaries who build autonomous, revenue-generating AI systems with zero overhead instead of funding bloated startups. One person, one vision, infinite scale.
Also: BCX
A new field dedicated to designing seamless, empathetic and efficient interactions specifically for non-human users like bots, agents and AI systems, because bots are proxies for human intent.
Dashboards are the comfort food of marketing: visually appealing snapshots of the past that give the illusion of control and answer questions you already asked, not the ones you should be asking. Exploration, not reporting, is the way out.
The shift from a traffic-chasing web to one of pervasive presence: AI curates and directs attention, brands win by earning AI trust through authoritative, structured data, and the website becomes the invisible data layer AIs tread on.
The point at which a measurement stand-in stops corresponding to the human it was built to represent. The number on the screen keeps moving, but what it stands for has come loose from the person it once tracked.
The amount of proof a decision deserves scales with the loss you take if the proof turns out to be wrong. Proof should be priced to the value at risk, the way an insurance premium is priced to the loss it covers.
A world where wealth and work are untethered from place: wealth goes mobile in borderless crypto wallets, work becomes location-agnostic, and people tied to one location risk being left behind.
Also: the Analog Backup
Keeping face-to-face connection, physical presence, local trust and offline skills and systems that still work when AI sits between us and every digital channel. Not against technology, but for humanity.
Also: time compression, five engines
The Singularity is reached when technological iteration cycles become shorter than human cognitive cycles and institutional cycles. It is driven by five compounding engines: AI, Robotics, Blockchain, Health and Energy.
Also: crowdsourced civilization
The safeguards that keep technological abundance free: radical cost transparency, ownership through participation, permissionless innovation, self-sovereign identity and data freedom, voluntary collective action (the crowdsourced civilization model) and rethinking education for flourishing.
Also: the Fifth Industrial Revolution
Martin's name for the Fifth Industrial Revolution. The Fourth was about building infrastructure; the Fifth is about extracting value from it and redefining what value means, starting with human-machine collaboration instead of replacement.
Also: the Velvet Cage, Velvet Tyranny
Deflationary abundance without safeguards: prices crash but power ascends, as mega-corps own the robot fleets, AI stays locked in proprietary vaults and essentials hide behind paywalls. Utopia with a leash.
No ideas match that search. Try another word, or .
Definitions are drawn from the articles themselves. New ideas are added as I write about them.
