Where else to learn AI honestly.

::host · Lex Fridman

Long-form (3-5hr) interviews with frontier researchers, CEOs, and adjacent thinkers; episode #490 'State of AI in 2026' with Lambert and Raschka is a representative current example. Mix of accessible framings and deep technical moments — value depends on the guest.

::host · Dwarkesh Patel

Highest-prep interviewer in the space. Episodes with Dario Amodei, Demis Hassabis, Sholto Douglas/Trenton Bricken, and Jensen Huang go into mechanics other shows skip. The interviewer reads the papers — it shows.

::host · swyx (Shawn Wang) and Alessio Fanelli

The AI Engineer beat. Coverage of agents, models, infra, and (since Jan 2026) a dedicated AI for Science sub-pod. Strong choice for builders who want to know what the labs are shipping and how production AI is actually wired.

::host · Nathan Labenz and Erik Torenberg

Biweekly interviews with builders and researchers at the frontier. Labenz was an early GPT-4 red-teamer and tends to ask the operationally-honest questions other hosts skate past.

::host · Tim Scarfe

Highest-rated technical AI pod on Spotify per their own data. Long, prep-heavy episodes on category theory, evolutionary algorithms, neuroscience-meets-ML, mech interp. Not background listening — bring a notebook.

::host · Sam Charrington

Long-running (since 2016) interview show covering ML/AI research, applied ML, MLOps, and increasingly agents and inference systems. Practical-engineering register — researchers, data scientists, ML engineers.

::host · Daniel Whitenack and Chris Benson

Changelog Media production focused on real-world AI implementation — orchestration, agents, infra, predictive models. Plain-spoken; good entry point for engineers who want signal without the founder-hype layer.

::host · Kevin Roose and Casey Newton (NYT)

Weekly NYT-produced show. Mainstream-accessible framing of the week in AI and tech, with reporter-style skepticism. Best on-ramp for people who don't want to read papers but need to track what's actually happening.

::host · Rob Wiblin, Luisa Rodriguez, Zershaaneh Qureshi

Long, thorough conversations on AI safety, alignment, biosecurity, and post-AI futures from the effective-altruism lens. 2026 has included episodes on RSI, AI biosecurity (VCT), and timeline updates.

::host · Karoly Zsolnai-Feher

Short, visually-driven walkthroughs of new papers in CG, ML, and AI. Over 1.77M subscribers as of 2026. Great onboarding format — 'what just shipped, and why should I care' in 3-5 minutes.

::host · Yannic Kilcher

The serious paper-walkthrough channel. Kilcher reads the paper end-to-end and pushes back on weak experimental setups instead of cheerleading. For people who want to learn how to critically evaluate ML research.

::host · Grant Sanderson

Visual deep-dives into linear algebra, calculus, and the math behind neural networks and transformers. Not novice-level despite the friendly art style — it's real math, taught honestly. The 'Neural Networks' and 'Transformers' series are field-defining.

::host · Andrej Karpathy

The 'Neural Networks: Zero to Hero' series builds backprop, makemore, micrograd, and a GPT from scratch in Python. Crossed 1M subscribers in early 2026. Karpathy joined Anthropic pretraining in May 2026 — back-catalog remains essential.

::host · Sam Witteveen

Google Developer Expert for ML. Hands-on tutorials on LangChain, agents, retrieval, function calling, and applied LLM patterns with companion GitHub code. Practical-builder register, not lecture-hall.

::host · Matthew Berman

5-6 videos a week. First-day coverage of new model releases with benchmarks and head-to-head comparisons. Particularly strong on open-source and local LLMs. Good for staying current on what's actually shipping.

::host · Allen Institute for Artificial Intelligence

Talks from AI2 researchers and visiting scientists — open models (OLMo), NLP, alignment, AI for science. Research-talk register, not produced for retention; valuable when you want the source instead of a recap.

::host · Jeremy Howard and Rachel Thomas

Free, code-first deep learning course. Lesson 1 deploys a working image classifier; the curriculum walks down to backprop, batch norm, ResNets, and transformers. Top-down teaching philosophy — build something working before you understand the internals.

::host · Hugging Face

Free, no ads. Transformers, Datasets, Tokenizers, Accelerate, the Hub, fine-tuning, and (in the 2026 expansion) building reasoning models. The canonical hands-on intro for working with open-weight LLMs.

::host · Andrew Ng / DeepLearning.AI

Weekly newsletter with Andrew Ng's letter at the top followed by curated coverage of research, hardware, business, and culture. Calm, signal-heavy, no hype-bait. Pairs well with DeepLearning.AI's short-course library.

::host · Jack Clark (Anthropic co-founder)

Weekly. Detailed analysis of cutting-edge papers, capability trends, and policy implications, plus a short fiction piece at the end of each issue. Crossed 100k+ subscribers in 2026. Clark writes from inside the frontier and tells you what to actually pay attention to.

::host · Ethan Mollick (Wharton)

Substack with 439k+ subscribers. Mollick writes about what AI actually changes for work and learning, with running hands-on experiments instead of vibes. The clearest non-engineer voice on how to use these tools well.

::host · Arvind Narayanan and Sayash Kapoor (Princeton)

Counterweight to the hype layer. Sharp on the difference between predictive AI (where most snake oil lives) and generative AI, on evaluation methodology, and on open-foundation-model risk framing. Required reading for anyone shipping AI claims.

::host · Nathan Lambert (Ai2)

1-3 posts per week from Ai2's post-training lead. Model reviews, RLHF/post-training internals, open-model ecosystem surveys. Inside-the-lab perspective without the marketing layer.

::host · Simon Willison

Coined 'prompt injection,' popularized 'AI slop' and 'agentic engineering.' Daily-ish notes on what's actually working with current models, plus deep dives on LLM tooling (Datasette, llm CLI, agentic patterns). Engineering register, no hype.

::host · Ben Thompson

The reference shop for tech strategy analysis. 2026 coverage includes 'Agents Over Bubbles,' the chip supply story, aggregators-and-AI, and the economic impact of agentic systems. Paywalled daily; free Monday article.

::host · Anthropic

First-party research from Anthropic on interpretability (translating Claude's thoughts to readable text, 2026), agentic misalignment reduction, constitutional classifiers, the Economic Index, and the Anthropic Institute. Source material, not commentary.

::host · OpenAI

First-party research and release notes. 2026 has included CoT-grading audits, the OpenAI Safety Fellowship, GPT-Rosalind for drug discovery, and the Privacy Filter open-weight PII model. Pair with the Alignment Research subsite.

::host · Rowan Cheung

2M+ subscribers in 2026. 5-minute daily of AI news, tools, and use cases. Tilts toward consumer-facing AI and tool launches — good first newsletter for non-technical readers; pair with a deeper feed if you're building.

::host · TLDR (Dan Ni)

Daily 5-minute Mon-Fri summary. Three sections: Big Tech & Startups, Science & Futuristic Tech, Programming/Design/Data Science. Free. Solid daily skim for ML engineers and PMs; not deep-dive material, but the digest is honestly curated.

::host · Noah Edelman and Pete Huang

Daily 700k+ subscriber newsletter, acquired by TechnologyAdvice in 2025. Friendly, lightly comedic register; explains why new releases matter for non-engineers. Best as a complement to a deeper source.