Unlikely guests, one recurring question: how do people learn?
Sources and Methods is a conversation show in which Alex and Matt question guests from fields that rarely overlap. A memoirist, a co-founder of fast.ai, an abacus math advocate and a professor of national security affairs all appear, and teaching comes up again and again across their conversations.
PRESS PLAY
Sources and Methods is a conversation show with two hosts, Alex and Matt, and a guest they want to understand from the inside. The guests don't share a field. What they tend to share is an interest in how people learn, and in who ends up knowing what. Eva Hagberg comes on to talk about 'How to Be Loved', a memoir about family and illness that the hosts call personal and that also takes in the health care system. The conversation turns to her time teaching architecture writing seminars at eight in the morning. She tried to make reluctant students care by starting from the personal, with daily journals about their rooms and drawn plans of their houses back home. She is sharp about the opposite reflex: the hand-wringing that decides nobody cares anymore and stops investing. Jeremy Howard, a co-founder of fast.ai, describes a project of subtraction. Five lines of code become three, then one, then possibly a button, because each line removed is a decision the user no longer has to make. The hosts praise fast.ai for taking deep learning beyond Silicon Valley insiders to domain experts in fields like medicine and history. In '36: Thomas Nichols', a professor of national security affairs makes a direct case about expertise. When citizens opt out of learning how things work, he argues, a small group of experts gets rewarded by default, not by conspiracy, and resentment builds between them and everyone else. He also names the book he rereads every year, C.S. Lewis's The Screwtape Letters. 'Learning the Abacus with RightLobeMath' turns to the soroban. The guest cites international math rankings, private abacus schools in India, Japan and China, and research on early math, and argues that this is one problem technology genuinely fits. Then there are the episodes without a guest. '35: Alex & Matt' wraps up 2016: the hosts name guests who intrigued them, catch up on their own work and joke that a listening milestone might just be spam bots. It also makes room for a more serious exchange about whether the lessons of international intervention in Afghanistan have been learned. In another catch-up, Alex mentions language coaching work in Jordan and a book of Taliban documents he published. The register is informal: one host will cut in on the other, and they admit when they are unsure of a detail. Show notes at sourcesandmethods.com link to the books and articles that come up. A listener drawn to technology could start with Jeremy Howard, and one more interested in civic life with Thomas Nichols.
Made for: Suits a listener who likes hearing people from unrelated fields explain how they think and work, and who is happy to jump from machine learning to memoir to national security between episodes.
What sets it apart: No single beat holds the guest list together, but education keeps coming back, from fast.ai courses and soroban math to writing seminars and the public's relationship with experts. The hosts also sometimes talk only to each other, as in a year-end episode that reaches Afghanistan.
Sources and Methods is a podcast hosted by Alex Strick and Matt Trevithick in which interesting people doing interesting things get to talk about the what, how and why of what they do.sourcesandmethods.com
An abacus meets AI for domain experts; your curiosity needn't pick a century.
Press play if you want
Short reviews from the PlayNext desk, based on the episodes we processed.
Eva Hagberg makes architecture and illness feel approachable through lived experience.
Eva Hagberg's How to Be Loved gives Sources and Methods a personal route into education, with illness opening onto family, friendship, health care, and the systems people move through while living with chronic conditions.
A clear, practical conversation about making deep learning more usable and learnable.
Deep learning is framed here as a matter of access: who gets to use it, how hard it is to learn, and what students need in order to work well. Jeremy Howard’s case is strongest when it stays practical.
Teaching talk with friction and nerve.
Héctor and Beto dissect Mexican grammar, slang, and news in native-speed Spanish conversations.
Small habit changes meet hands-on AI experiments.
["Eva Hagberg says she has \"three to 17 concurrent lives and careers\" and uses a UC Berkeley PhD in visual and narrative culture to connect architecture, memoir, and criticism.
["Sources and Methods talks with Jeremy Howard about deep learning, fast.ai, and why he wants AI to be accessible to domain experts.", "Matt and Alex host Howard, who says his path ran through McKinsey & Company, Fastmail, insurance pricing, Analytic, and Kaggle.", "He argues that machine learning should be taught from real tasks first, using whole-game teaching, transfer learning, and a top-down course design.", "He also discusses Chinese study, Anki, Pop-Up Chinese, exercise, and note-taking as habits that changed how he learns."]
Alex and Matt interview Stephen Tankel about his book on US counterterrorism partnerships and the complexities of cooperating with other nations.
A data scientist discusses using code and data to address justice issues, including monitoring border wall events and dealing with burnout.
Interview with Chris Lee of Launch School discussing their slow, fundamentals-first approach to teaching coding, contrasting with typical bootcamps.
Hosts Alex and Matt return for season three, catching up on their recent projects including a Taliban reader and a new company launch.
Thomas Nichols connects the decline of expertise to internet search, confirmation bias, and what he calls narcissistic overconfidence.
Alex's CoachBot turns 385 language-learning tasks into time-based prompts for studying when motivation or energy is low.