Frameworks
Useful Ways To Think Before You Reach For The Tool
Reusable mental models for judging AI by finished work, understanding what happens when AI sits between people and the web, and following where economic value moves as old constraints disappear.
The Effective Task Cost Framework
Don't judge an AI tool by the price of a single reply. Judge it by what it actually costs you to get a real task to a finished, usable result. The comparison that matters isn't "cheap model versus expensive model" — it's "cheap finished work versus expensive finished work." A tool that answers instantly but needs five follow-up corrections can end up costing you more time than a slower tool that gets it right on the first try.
Task Economics
Where Does The Job Actually End?
Two frameworks for telling the difference between "the AI produced something" and "the work is actually done."
The Completion Boundary
An AI finishing its reply is a technical event. The job being done is a human judgement call. Say you ask an AI to build a travel itinerary — it stops the moment it's written everything out, but the itinerary isn't actually "done" until the flight times are correct, the hotel is bookable, and the budget adds up. Decide what "done" looks like for your specific task before you judge how good or fast a tool was.
Consumption vs. Completion
The number of messages, tokens, or minutes you used tells you how much you consumed — it doesn't tell you whether the task succeeded. If it takes you six back-and-forth messages to get a usable spreadsheet formula, that's six units of consumption for one completed job. When judging whether a tool is worth using, count the finished results you actually got, not how much you typed or how long you waited.
Interface
What Changes When The AI Becomes The Middleman?
Two frameworks for thinking about what happens to websites and businesses once people start going through an AI assistant instead of going directly to a site.
The AI-Mediated Web
The old pattern was simple: you go to a website, you look around, you decide, you buy. Increasingly, a new pattern is showing up: you ask an AI, it searches, compares, and sometimes even books or buys on your behalf, and the website only shows up at the very end, if at all. For example, instead of browsing five hotel booking sites yourself, you might ask an AI to find and book the best option — the hotel's website still exists and still matters, but it no longer gets your attention or influences your decision the way it used to.
The Unbundling of the Page Visit
A single visit to a website used to do five jobs at once: you discovered something, you paid attention to it, you saw ads, the site collected data about you, and maybe you bought something — all in one visit. When an AI handles the search and comparison for you, those five jobs can happen separately, or some might not happen at all. If an AI recommends a product and you buy it through the AI's link, the seller might get the sale but never got your attention, never showed you an ad, and never collected your browsing data the way they used to. The useful question is: of those five things a page visit used to bundle together, which ones still require you to show up in person, which ones the AI can now do for you, and which ones simply stop happening?
Market
If Attention Stops Paying The Bill, What Does?
Two frameworks for thinking about how businesses make money once people stop personally browsing, clicking ads, and paying attention the old way.
Attention Fades. Money Moves.
A lot of the internet is funded by advertising, which depends on people actually looking at pages and ads. If AI assistants start doing the browsing and comparing for people, that kind of attention becomes less reliable to sell. Money doesn't just vanish, though — it tends to move towards things that are still worth paying for even when nobody's staring at a screen: reliable up-to-date information, results that are verifiably correct, being the option an AI recommends, or controlling the final step where someone actually completes a purchase or booking. This isn't a prediction of one winner — it's a map of where the money can flow once the old advertising model weakens.
Follow The Scarcity
When AI makes something abundant and easy — like writing, searching, or comparing options — that doesn't mean money disappears from that space. It just means the money moves towards whatever is still hard to get or hard to fake. Right now, the things becoming more valuable, precisely because AI can't easily replace them, include live and current information, results you can actually verify are correct, user trust, control over someone's identity or login, permission to act on someone's behalf, and the relationship with the actual paying customer. When evaluating any AI-driven business idea, ask: once AI makes the obvious part cheap and easy, what's the part that's still genuinely scarce — and can you own that part?