Everything referenced in the chapter, grouped by how central it is, with a short note on what each one is for.
Prices and free tiers change often, so check a resource's own page before you plan around one.
Browse: Start here · Optional depth
Start here
The resources on the main path for this chapter, in the order the chapter uses them. If you only do a few things, do these.
How to Interview and Hire ML/AI Engineers
Eugene Yan · Article · Free · 1 hour
One hiring manager's detailed rubric: software engineering, data literacy, uncertainty, evaluation, and scientific breadth/depth, plus behavioural signs of handling ambiguity and execution. Use it to review your project and prepare concrete examples. It reflects one manager's priorities, so check each employer's actual interview format.
The Rise of the AI Engineer
Shawn "swyx" Wang · Article · Free · 1 hour
The 2023 essay that popularised this role's current framing: developers building products on pretrained models, with substantial work in integration, evaluation and operations. It explains the work, not how many jobs exist or whether an employer requires a degree. Its forecasts are the author's argument, not measured hiring results.
Worth knowing. Read it as the origin of the term, not current market analysis. Its 2023 supply/demand reasoning predates both the agent era and the shrinking junior market.
AI and Job Postings: From Destruction to Creation?
Guillermo Gallacher · Article · Free · under an hour
An analysis of US Indeed postings, published July 2026. Software development postings rose roughly 15% from late February 2025 while overall postings fell 7%. They were still 27.5% below February 2020. In May 2026 senior roles made up 71% of the year-over-year increase and AI-titled roles 37%. These categories can overlap. Cite the measured market, dates and denominators with the findings. Postings and a rebound do not prove that AI caused hiring, or that your country follows suit.
Worth knowing. The headline finding is a correlation. The roughly 15% rebound is measured from a coding-assistant release date, which invites a causal reading the data does not support.
Teach Yourself Programming in Ten Years
Peter Norvig · Article · Free · under an hour
A lasting argument for steady practice: write programs, read other people's code, build projects, ask for feedback and understand the hardware. The title is a reply to instant-expertise marketing, not a tested ten-year rule for every learner or role.
uv: Working on projects and locking dependencies
Astral maintainers · Documentation · Free · 1 hour
Start with installation, then use this
for the Chapter 1 setup, interpreter choice and project files. Read
locking and syncing to understand --locked,
planned upgrades and syncing the environment. Commit the Python version and dependency lock, then
check a fresh clone. A lockfile does not capture datasets or hardware.
Official compute limits and pricing
Service maintainers · Documentation · Free to read · 30 minutes
Start with the Colab FAQ linked in the heading, then check Kaggle notebook documentation, Lightning pricing and Modal pricing before choosing a runtime. Colab does not promise a specific free GPU or a fixed, published usage limit. Modal Starter credits do not cover per-token Shared Endpoints charges. Choose by measured memory and runtime needs. Save checkpoints outside temporary sessions, and set a spending limit before paid use. Chapters 1–3 can run on CPU.
Optional depth
Worth your time if the chapter left you wanting more, or if this is where you want to specialise.
2026 Stack Overflow Developer Survey
Stack Overflow research team · Dataset · Free · 1 hour
The latest edition was released on October 6, 2026. Explore AI use, trust, work and learning, and check the respondent count and subgroup for each question. It is a voluntary survey, so reported tool use and opinions do not measure productivity or show what causes better learning. Before comparing with the 2025 AI section, check whether the questions and respondents changed.
Worth knowing. A percentage of AI users is not automatically a percentage of all developers. Each year's respondents are a new sample, not a panel followed over time.
Experiments on AI assistance and learning
Zara Contractor & Germán Reyes · Preprint revised 4 October 2026 · Free · 1–2 hours of selected reading
In an AI-access experiment, 211 undergraduates wrote an essay on an unfamiliar topic. Access improved their later unaided knowledge-test scores, with a smaller gain at one week. Its tutoring and delegation comparisons are descriptive, and an essay differs from learning a programming library. Pair it with Heickal and Lan's 2026 classroom-feedback workshop paper. There, feedback went to failed submissions, and the reported links are with completion and retries, not later unaided mastery. Read both alongside the book's coding trial to keep interventions and outcomes apart. A polished artefact made with AI help is not a retention test.
Deliberate Practice and Performance: A Meta-Analysis
Brooke N. Macnamara, David Z. Hambrick & Frederick L. Oswald · Paper · Abstract free; full text may require access · 2 hours
Read it for one distinction: practice can help without explaining every difference in performance. The 2014 meta-analysis found large variation across domains, and its 2018 correction revised some estimates. How each study defines and measures deliberate practice matters. Variance explained across people does not set your own learning ceiling or show that feedback is unnecessary.
Worth knowing. This is older research with a published correction, not a current experiment on learning AI engineering.
Make It Stick: The Science of Successful Learning
Peter C. Brown with Henry L. Roediger III & Mark A. McDaniel · Book · Paid · 8–10 hours
A readable account of research on learning and memory, from a ten-year collaboration among 11 cognitive scientists at six universities. Its focus on retrieval practice, spacing and interleaving fits this roadmap: rebuild a small example from memory, revisit it later, and check your mistakes against working code. The book is optional. Apply the practices to your weekly sessions before adding more reading.
The MOOC Pivot
Justin Reich & José A. Ruipérez-Valiente · Science paper, author postprint · Free · 1 hour
The original 2019 paper using MITx/HarvardX data from 2012–2018, replacing a press summary. It separates completion for all participants, those who said they intended to finish and paid verified-track learners. Many registrants never opened the courseware. So overall completion says little about a committed learner, and it does not mean a certificate only signals enrolment. This book's practical answer: make each week produce working code, and ask for feedback.
Worth knowing. Older platform data is useful context, not a current estimate for every online course or learner.
The 2026 AI Index Report — Economy chapter
Stanford Institute for Human-Centered AI · Paper · Free · 2 hours
An annual summary of investment, adoption, labour market and productivity research. Before reusing a number, follow its citation to the original study. Employer expectations, job postings, employment and measured task productivity answer different questions. Check the population, period and causal design, especially when applying a US finding to your own market. Treat the chapter as a map of sources, not one experiment with a common denominator.
The Labor Market Is Tilting Toward Seniority
Felix Aidala & Sneha Puri, Indeed Hiring Lab · Article · Free · under an hour
A second Indeed analysis: in Q1 2026, senior roles made up 69.3% of US software development postings and entry-level roles 4.5%. Across all US postings, senior roles were roughly 14% of the total. Note how the authors classify seniority, and keep a share separate from a change over time. Use it to see what experience your target vacancies ask for and what you still need to prove.
Worth knowing. It counts US postings on one job site, not hires, so it shows demand rather than who actually gets the jobs. The authors also weigh causes other than AI, such as remote work and interest rates.