Journey · 11 steps · 10 sources
Judging AI as an investment or as a tool starts with how a model actually reads text. This path goes from tokens and attention to evaluation, then to the shift in what the hardware runs, the single points of failure in the supply chain and the return the spending has to earn, and ends with what changes in the work itself.
Reason about a model in tokens: that is what it reads, what fills its window and what you pay for.
VideoMost Devs Don’t Understand How LLM Tokens Work· Matt Pocock
A long window degrades before it fills; curate and order what goes in rather than relying on size.
VideoWhy LLMs Get Dumb (Context Windows Explained)· NetworkChuck
Write to a model as you would brief a capable junior: the task, their role, the context they lack, and what the output should look like.
VideoPrompting 101· Anthropic
Build the test set before the feature, measure every change against it, and keep a person in the loop where the metric cannot be trusted.
VideoHow to Systematically Set Up LLM Evals· Dave Ebbelaar
Decide which mistake is more expensive before tuning a model; the threshold is a business choice the statistics only express.
VideoPrecision vs Recall, With a Clear Example· Kimberly Fessel
A compute vendor's case is a flywheel from agents to tokens to compute; the buyer's case is the return on that compute, and both have to be read together.
LetterNVIDIA's fiscal 2026 letter from Jensen Huang· Jensen Huang
Find the narrowest step in a supply chain; that is where the risk, the pricing power and the politics all sit.
BookChip War· Chris Miller
Spending is only growth if the return on it exceeds the cost; estimate the return under several assumptions before admiring the spend.
LetterHayden Capital's Q2 2026 letter, on the "SaaS-pocalypse" and returns on reinvestment· Fred Liu
In any industry, ask who holds the data the models need; the shift favours them more than the vendors of the models.
LetterCapital One's 2024 letter from Rich Fairbank· Rich Fairbank
Prepare for the parts of the job the tools make scarce, judgement and client sense, rather than the parts they make cheap.
VideoGoldman Sachs CEO David Solomon on AI, Debt and America’s Future· Prof G Markets
Name the purpose of a job separately from its tasks; automation changes the tasks and leaves the purpose to be done better or worse.
LetterNVIDIA's fiscal 2026 letter from Jensen Huang· Jensen Huang