How language models work, what they cost, and what they change in finance and engineering jobs.
Reason about a model in tokens: that is what it reads, what fills its window and what you pay for.
Explaining how a language model works, estimating the cost of an AI feature, and debugging a model that fails at a character-level task.
From Most Devs Don’t Understand How LLM Tokens Work · Matt Pocock
A model predicts the past decisions in its training data; if those were biased, the model is a bias machine with a straight face.
Any model used on people, credit underwriting, hiring tools, and AI governance questions in a bank.
From Weapons of Math Destruction · Cathy O’Neil
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.
Using any model on real work, and explaining in an interview how you would get a useful answer from one.
From Prompting 101 · Anthropic
Separate what you can do from the structure that makes it valuable, and watch the structure.
Choosing a seat, judging the effect of automation on a role, or answering what AI changes about an analyst job.
From Floored · 2009
Whatever you can measure and reward will be maximised literally, including the gaps between the measure and the goal.
AI product and risk discussions, designing any metric or bonus scheme, and evaluating a model's objective.
From The Alignment Problem · Brian Christian
Name the purpose of a job separately from its tasks; automation changes the tasks and leaves the purpose to be done better or worse.
Answering what AI changes about an analyst role, planning a career, or assessing headcount claims in a bank.
From NVIDIA's fiscal 2026 letter from Jensen Huang · Jensen Huang
Spending is only growth if the return on it exceeds the cost; estimate the return under several assumptions before admiring the spend.
Any capital-intensive growth story, the AI capital expenditure debate, and valuation of a company reinvesting most of its cash flow.
From Hayden Capital's Q2 2026 letter, on the "SaaS-pocalypse" and returns on reinvestment · Fred Liu
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.
Semiconductor and data-centre coverage, technology investing, and any debate about AI capital spending.
From NVIDIA's fiscal 2026 letter from Jensen Huang · Jensen Huang
A long window degrades before it fills; curate and order what goes in rather than relying on size.
Designing any system that feeds documents to a model, and explaining why a model missed something that was in its context.
From Why LLMs Get Dumb (Context Windows Explained) · NetworkChuck
In any industry, ask who holds the data the models need; the shift favours them more than the vendors of the models.
Bank and fintech coverage, the question of who wins from AI in finance, and any moat built on data.
From Capital One's 2024 letter from Rich Fairbank · Rich Fairbank
Institutional AI is scoped, governed and data-limited; expect to work inside those limits and to be able to explain them.
Questions about AI in banking, technology roles inside financial firms, and any claim about what AI is doing to the industry.
From Goldman CIO Marco Argenti on the Warp-Speed Improvements in AI · Bloomberg Odd Lots
Decide which mistake is more expensive before tuning a model; the threshold is a business choice the statistics only express.
Any model used for screening, fraud, credit or hiring, and the interview question of how you would evaluate a classifier.
From Precision vs Recall, With a Clear Example · Kimberly Fessel
Prepare for the parts of the job the tools make scarce, judgement and client sense, rather than the parts they make cheap.
The interview question about what AI changes for an analyst, and planning what to get good at before you start.
From Goldman Sachs CEO David Solomon on AI, Debt and America’s Future · Prof G Markets
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.
Any AI product discussion, any system design that includes a model, and the question of how you know an AI feature works.
From How to Systematically Set Up LLM Evals · Dave Ebbelaar