AI & software podcasts worth following — 2026

TL;DR — pick by intent, not by hype.

Building with AI: Latent Space [1] + Practical AI [3] + Cognitive Revolution [11].

Tracking research / open models: Interconnects [2] + ML Street Talk [5] + Dwarkesh [4].

Daily news without scrolling X: AI Daily Brief [6] + Hard Fork [7] + ChinaTalk [18].

Software craft: The Pragmatic Engineer [8] + The Changelog [9] + SE Radio [22].

Capital & strategy: BG2 [25] + Acquired [10] + Sharp Tech / Dithering [20].

If you only have one slot: Latent Space for builders, Dwarkesh for thinkers, The Pragmatic Engineer for working software engineers.

AI engineering — building with the tools

Listen here when you write code that calls models or ships agents. Practitioner-density per minute is the metric that matters.

Show	Hosts	Cadence	Why high-signal
Latent Space [1]	swyx (Shawn Wang), Alessio Fanelli	Weekly + bonuses	Tightest coverage of what AI labs and tooling startups actually ship — agents, eval harnesses, training infra. April 2026 alone: Shopify’s AI rollout, Notion’s knowledge agents, OpenAI Frontier’s “extreme harness engineering” [1].
Practical AI (Changelog) [3]	Daniel Whitenack, Chris Benson	Weekly	Implementation-first lens. April 2026: agentic coding economics, Comma AI self-driving, edge-AI constraints, Anthropic Claude Code leak post-mortem [3].
The Cognitive Revolution [11]	Nathan Labenz, Erik Torenberg	Biweekly	Long, technical conversations with builders + researchers. Best mid-year and end-of-year recap series in the genre [11].
No Priors [12]	Sarah Guo, Elad Gil	Weekly	VC lens — useful for who’s getting funded, what investors think will matter next [12].
TWIML AI Podcast [13]	Sam Charrington	Weekly	Broad ML coverage stretching back to 2016. Recent: production-grade diffusion LLMs, multi-agent systems at Capital One, reasoning in small models [13].

Heads-up: swyx confirmed in early 2026 that Latent Space is becoming a podcast network — first companion show is AI for Science, plus AINews (80k subs) integration for daily updates [14]. If you only follow one AI eng feed, this one is now structurally widening its coverage.

AI research & long-form thinking

Depth over breadth. These are the shows you put on for a flight, not a commute.

Show	Format	Why
Dwarkesh Podcast [4]	1–3h interviews	Deeply-researched interviews with AI researchers (Karpathy, Sutskever), labs (Dario Amodei), and CEOs. The April 2026 Jensen Huang episode covered TPU competition, US chip controls on China, NVIDIA’s supply-chain moat, and why NVIDIA isn’t a hyperscaler [15].
Machine Learning Street Talk [5]	1–3h, technical	Tim Scarfe (PhD ML, ex-Microsoft Principal Engineer, ex-bp Chief Data Scientist). The most technically rigorous AI podcast running — alignment, mechanistic interp, scaling-law internals. Highest-rated technical AI podcast on Spotify [5]. ⚠ Steep — assumes you’ve read the papers.
Interconnects [2]	Essays + interviews	Nathan Lambert, post-training lead at Ai2 [16]. Clearest signal on the open-weights ecosystem — RLHF, RewardBench, Olmo, Tülu. Listen here if you care about non-frontier-lab models.
Lex Fridman Podcast [17]	2–4h interviews	Inconsistent for AI specifically, but episode #490 (“State of AI in 2026” with Nathan Lambert + Sebastian Raschka) is the cleanest 2026 overview of pre/mid/post-training, AGI timelines, and AI compute infra in any medium [17].

AI news & policy — staying current without scrolling

Show	Cadence	Use it for
The AI Daily Brief [6]	Daily, 15–30 min	Nathaniel Whittemore’s morning briefing — highest-signal daily AI summary; 960+ episodes since April 2023 [6].
Hard Fork (NYT) [7]	Weekly	Casey Newton + Kevin Roose. Mainstream-press frame on AI/policy. Good for: the consensus narrative. Bad for: depth or contrarian takes [7].
ChinaTalk [18]	2–3x/week	Jordan Schneider on China + tech + US-China policy. 65k substack subs as of late 2025; growing defense focus including a new weekly Second Breakfast show [19]. The default for “what’s actually happening in Chinese AI labs”.
Dithering [20]	2x/week, 15 min	Ben Thompson + John Gruber, exactly 15 minutes, one tech topic. Highest information density per minute on this list. Recent: Jensen Huang’s GTC 2026 keynote, “LLM paradigm changes” [20].

Software engineering — craft and big-tech reality

Show	Hosts	Why
The Pragmatic Engineer [8]	Gergely Orosz	Big Tech and startup engineering from the inside. Recent / upcoming guests: Thuan Pham (ex-Uber CTO, now Faire CTO), Martin Kleppmann, DHH [8].
The Changelog [9]	Adam Stacoviak (Jerod Santo retired March 2026 [27])	Open source weekly — interviews with maintainers and founders. March 2026: Tailscale’s TSIDP/Aperture roadmap, GitHub Copilot’s Burke Holland on agentic dev workflows [9]. ⚠ Format in transition post-Jerod.
Software Engineering Daily [21]	Various (Gregor Vand, Sean Falconer for SED News)	3 episodes/week, 2,100+ episodes since 2015. Broad — cloud, AI, developer culture, Silicon Valley. Monthly SED News recap is a good catch-up format [21].
Software Engineering Radio (IEEE) [22]	Various	Slow burn (~2/month), tutorial depth on a single topic — the antidote to news-bait podcasts. Recent: Martin Kleppmann on local-first software (Apr 15), Marc Brooker on spec-driven AI dev (Mar 4) [22].
Lenny’s Podcast [23]	Lenny Rachitsky	Product / growth crossover. Recent AI episodes punch above weight: Boris Cherny on how Claude Code grew to 4% of public GitHub commits [23]; Simon Willison’s April 2026 episode argued Nov 2025 was the inflection where AI coding agents crossed “mostly works → actually works” [24].

Capital, strategy, and company arcs

Show	Hosts	Why
Acquired [10]	Ben Gilbert, David Rosenthal	4-hour deep dives on company histories. #1 tech podcast on Apple/Spotify, >1M listeners per episode [10]. The right format when you want to understand how a category was built — NVIDIA, TSMC, Amazon. ⚠ Long; treat each episode as a half-day commitment.
BG2 [25]	Brad Gerstner (Altimeter), Bill Gurley	Biweekly markets/AI/tech. 2026 episodes featured Jensen Huang on the OpenAI $100B partnership and OpenAI’s Nick Turley on ChatGPT-as-super-assistant [25]. Closest you’ll get to listening in on a Sand Hill Road dinner.
Sharp Tech [26]	Andrew Sharp, Ben Thompson	Twice-weekly Stratechery-aligned analysis. Recent: TSMC pricing power, Meta CapEx, OpenAI’s enterprise pivot, NVIDIA’s inference pivot [26]. Pair with Dithering for short + long Ben Thompson.

Pick-by-time-budget

Listening budget	Stack
< 1 hr/week	AI Daily Brief (skim) + 1 Dithering.
2–3 hrs/week	+ Latent Space + The Pragmatic Engineer.
5+ hrs/week	+ Dwarkesh / MLST (alternating) + Interconnects + Cognitive Revolution.
Audiobook-replacement	Acquired’s NVIDIA episodes [10] and Dwarkesh’s longer interviews [4].

Notes on what’s not on the list

Generalist tech-celebrity interviews that don’t add information beyond what a guest already posts on X — skipped.
Vendor podcasts disguised as content. A handful of enterprise-AI shows are sales channels with light editorial polish; listen with that lens or skip.
Daily news bulletins beyond AI Daily Brief. Marginal returns drop fast — one daily AI brief plus a weekly recap (Hard Fork or Practical AI) covers most listeners.
Sharp China, Greatest of All Talk and other Stratechery-network spinoffs [20] are excellent if you’ve already maxed out on Sharp Tech and Dithering — tier 2 only because they’re behind the paywall.