9 Best AI Podcasts in 2026 (Verified Still Publishing)
Short answer: the Lex Fridman Podcast for the widest range of guests across AI, science and culture; Latent Space if you build with AI and want engineering depth; Dwarkesh Podcast for the most researched long interviews with frontier AI researchers; and The AI Daily Brief if you want a daily habit instead of a weekly one. All nine picks below are free to listen to, and I checked each one's own feed on September 24, 2026 to confirm it is still publishing.
I read AI newsletters for the fast version of the news. I put on an AI podcast when I want the argument behind the headline, the part a 200-word newsletter blurb cuts out. That distinction matters because the AI podcast space has thinned out this year: Hard Fork, the New York Times show that a lot of "best AI podcast" lists still lead with, wrapped its run in September 2026, and hosts Kevin Roose and Casey Newton moved to a new video-first show, Machine Gods, on NPR. I dropped it from this list and checked every remaining pick's feed the same day I wrote this.
Dupple data: Toolradar, our software directory, tracks 5 podcasting tools such as Acast, RSS.com and Zencastr as of September 2026, for readers who want to launch a show rather than just listen to one; see the best podcast hosting platforms if that's you.
Quick comparison
| Podcast | Host(s) | Network | Cadence | Best for |
|---|---|---|---|---|
| Lex Fridman Podcast | Lex Fridman | Independent | Roughly every 1-2 weeks, listed "Updated Monthly" on Apple | Long-form interviews across AI, science and culture |
| Latent Space | swyx and Alessio | Independent (Substack) | Weekly | AI engineers who want build-level detail |
| Dwarkesh Podcast | Dwarkesh Patel | Independent (Substack) | Weekly | Deeply researched interviews with frontier researchers |
| No Priors | Sarah Guo and Elad Gil | Conviction | Weekly | A VC and founder's read on where AI is going |
| The Cognitive Revolution | Nathan Labenz and Erik Torenberg | Turpentine | Several times a week | Frequent builder interviews plus a news roundup |
| The AI Daily Brief | Nathaniel Whittemore (NLW) | Turpentine | Daily | A short daily AI news habit |
| TWIML AI Podcast | Sam Charrington | Independent (CloudPulse Strategies) | Weekly | Long-running, research-heavy ML interviews |
| Practical AI | Daniel Whitenack and Chris Benson | Changelog Media | Weekly | Engineers shipping AI in production |
| Eye on AI | Craig S. Smith | Independent | Roughly every two weeks | A veteran journalist's take on AI research and policy |
What makes an AI podcast worth your feed
A podcast earns a spot on your phone if it still publishes on a schedule you can rely on, if the host asks questions a newsletter blurb can't answer, and if the guest list is people actually building or researching the thing, not just commenting on it. Several long-running AI shows went quiet or folded in 2026, so the first filter below is: does the feed have a new episode in the last two weeks.
How we chose
I checked each show's own Apple Podcasts listing and, where it had one, its own website on September 24, 2026, for host names, publishing cadence and the date of the newest episode. Nine made the cut; Hard Fork did not, because it stopped publishing that same month. No podcast paid for inclusion, and Dupple runs its own newsletter (Techpresso), which is disclosed on our about page rather than ranked here.
Lex Fridman Podcast, the widest net for long interviews
Lex Fridman is an MIT research scientist who has been running long-form interviews since 2018, and AI is only part of the beat: recent episodes have covered psychiatry, programming and MMA alongside frontier AI labs. Apple Podcasts lists the show's cadence as "Updated Monthly," but episode numbering (episode 502 landed September 17, 2026) shows new interviews arriving closer to every one to two weeks in practice.
listeners who want AI covered alongside science, history and culture in one feed, not an AI-only show.
Standout: three-hour, unhurried conversations that let a guest actually finish a technical explanation.
Watch out: episodes run two to four hours and the schedule is genuinely irregular, so it doesn't work as a commute habit.
Verdict: the default pick if you want range over frequency.
Latent Space, the engineering-depth pick
Latent Space is hosted by swyx (Shawn Wang) and Alessio, published weekly through their own Substack, which lists over 201,000 subscribers. A recent episode dug into AI biosecurity with the CEO of Radical Numerics, the kind of guest a general-interest show wouldn't book.
people who build AI products and want the conversation to go past the demo into infrastructure and model internals.
Standout: guests are practitioners at labs and infra companies, not pundits.
Watch out: the show assumes ML background; it does not stop to define terms for a general audience. Latent Space also runs a paid Substack tier, but the subscribe page does not publish its price.
Verdict: the pick if Lex Fridman's range feels too broad and you want engineers talking to engineers.
Dwarkesh Podcast, the most researched interviews
Dwarkesh Patel publishes weekly and has built a reputation for spending weeks preparing a single interview. His Substack lists over 109,000 subscribers. The most recent episode at the time of writing paired him with OpenAI's Noam Brown on agent swarms and recursive self-improvement.
listeners who want a host who has read the guest's papers before asking the question.
Standout: interviews run 60-90 minutes and stay on one thread instead of touring five topics.
Watch out: the prep-heavy format means fewer episodes than a daily show, and the density can be a lot for a first listen.
Verdict: the pick when you want one deep conversation a week over five shallow ones.
No Priors, the venture capital lens
No Priors is co-hosted by Sarah Guo, founder of the investment firm Conviction, and Elad Gil, and publishes weekly. A recent episode interviewed the CEO of Sequence Holdings on rebuilding incumbent companies for the AI era.
founders and operators who want AI news framed around what it means for a startup's roadmap and fundraising, not just the research.
Standout: both hosts sit on cap tables, so questions about business model and moat land differently than from a journalist.
Watch out: the VC framing means less raw technical depth than Dwarkesh or Latent Space; treat it as the business layer, not the research layer.
Verdict: add it if your job is deciding what to build or fund, not what to train.
The Cognitive Revolution, the high-frequency builder feed
The Cognitive Revolution is hosted by Nathan Labenz and Erik Torenberg on the Turpentine network. Apple Podcasts' own show description calls it "a biweekly podcast," but the feed actually publishes several times a week, mixing full guest interviews with a shorter "AI:AM" news roundup format.
listeners who want more than one AI podcast episode a week without adding a second show to the feed.
Standout: the AI:AM segments work like a spoken newsletter between the longer interviews.
Watch out: the volume is real. Skipping the AI:AM episodes and keeping only the guest interviews cuts the time commitment roughly in half.
Verdict: the pick if The AI Daily Brief feels too short and Dwarkesh feels too infrequent.
The AI Daily Brief, the daily habit
The AI Daily Brief is hosted by Nathaniel Whittemore, known as NLW, also on the Turpentine network, and publishes daily. It's built for a five-to-fifteen-minute listen, not a deep dive.
anyone who wants an AI news habit at the same slot every day, similar to how Techpresso works for a five-minute morning read.
Standout: an ad-free Patreon tier is priced at $3/mo, a rare case of an AI podcast publishing an actual subscription price rather than staying ad-supported only.
Watch out: daily cadence means daily volume; a few days off and the backlog stacks up fast.
Verdict: pair it with one weekly deep-dive show rather than using it alone.
TWIML AI Podcast, the long-running research interview
The TWIML AI Podcast is hosted by Sam Charrington, who started the show in 2016 under his own company, CloudPulse Strategies. It publishes weekly and has passed 7 million total downloads; the most recent episode at the time of writing was numbered 777.
listeners who want a research-literate host interviewing ML practitioners and academics, not just lab PR teams.
Standout: nearly a decade of past episodes means you can trace how a research area evolved over time.
Watch out: the tone skews academic; it's a slower listen than the daily or VC-framed shows on this list.
Verdict: the pick for ML practitioners who want depth over headlines.
Practical AI, the production-engineering show
Practical AI is co-hosted by Daniel Whitenack and Chris Benson, produced by Changelog Media, and has run weekly since 2018. The most recent episode, number 373, published September 24, 2026, covered moving AI agents from prototypes into enterprise deployment.
engineers who need MLOps and deployment detail, not trend commentary.
Standout: guests are practitioners shipping AI systems, which keeps the conversation closer to "what broke" than "what's next."
Watch out: the guest list leans toward working engineers over public figures, so it's a quieter, less star-driven listen than Lex Fridman or No Priors.
Verdict: the pick if your day job is keeping an AI system running, not evaluating which lab to trust.
Eye on AI, the veteran journalist's angle
Eye on AI is hosted by Craig S. Smith, a longtime New York Times correspondent, and publishes roughly every two weeks. Episode 382, the most recent at the time of writing, covered autonomous drone targeting systems in Ukraine.
listeners who want AI research and policy framed with the instincts of a career journalist rather than an operator or investor.
Standout: the reporting instinct shows in story selection, recent episodes have covered military and geopolitical uses of AI that the builder-focused shows skip.
Watch out: the every-two-weeks cadence means it won't replace a daily or weekly habit on its own.
Verdict: add it for the stories the Silicon Valley-centric shows on this list don't cover.
How to choose
You want one show and nothing else: Lex Fridman Podcast. The range covers AI, science and culture, even if the schedule is irregular.
You build AI products for a living: Latent Space plus Practical AI. Engineering detail and production reality, back to back.
You want the deepest single interview each week: Dwarkesh Podcast. Nothing else on this list prepares as hard for one conversation.
You make funding or roadmap calls: No Priors. The VC framing answers "should we build this" better than a research-first show does.
You want news daily, not weekly: The AI Daily Brief, paired with The Cognitive Revolution's AI:AM segments if one daily show isn't enough.
You're a working ML practitioner: TWIML AI Podcast. Nearly a decade of research-literate interviews.
You want the story angle newsletters miss: Eye on AI, for the policy and geopolitical framing.
FAQ
What are the best AI podcasts to follow in 2026?
The Lex Fridman Podcast, Latent Space and Dwarkesh Podcast lead for range, engineering depth and research rigor, respectively. No Priors covers the venture angle, The AI Daily Brief and The Cognitive Revolution cover daily and high-frequency news, and TWIML AI, Practical AI and Eye on AI round out research, production and policy angles.
What happened to Hard Fork?
Hard Fork, the New York Times AI podcast hosted by Kevin Roose and Casey Newton, published its final episode in September 2026, closing what the hosts called "this chapter" of the show. Roose and Newton moved to a new video-first show, Machine Gods, launching on NPR. Any "best AI podcast" list still recommending Hard Fork as an active show is out of date.
Is there a free AI podcast worth subscribing to?
Yes, all nine podcasts on this list are free to listen to on the standard podcast apps. The AI Daily Brief also sells a $3/mo ad-free Patreon tier, and Latent Space runs a paid Substack tier without a published price, but neither requires payment to hear every episode.
How often do these AI podcasts publish new episodes?
The AI Daily Brief is daily. Latent Space, Dwarkesh Podcast, No Priors, TWIML AI Podcast and Practical AI are weekly. The Cognitive Revolution publishes several times a week. Eye on AI publishes roughly every two weeks, and the Lex Fridman Podcast is the least predictable, landing closer to every one to two weeks despite Apple's "Updated Monthly" listing.
Which AI podcast is best for beginners versus practitioners?
The Lex Fridman Podcast and The AI Daily Brief require the least background because both hosts explain context as they go. Latent Space, Dwarkesh Podcast and TWIML AI Podcast assume you already know the vocabulary and move faster.
Do any AI podcasts have a paid tier?
Most don't. The AI Daily Brief's Patreon is the only one on this list with a published USD price ($3/mo, ad-free). Latent Space offers a paid Substack tier but does not list its price on the subscribe page.
Which AI podcast covers AI research most deeply?
Dwarkesh Podcast and TWIML AI Podcast go deepest into research, with Dwarkesh favoring single, heavily prepared interviews and TWIML favoring a wider rotation of academic and industry guests across nearly a decade of episodes.
Bottom line
Start with the Lex Fridman Podcast if you want one AI podcast that covers the widest ground, add Latent Space or Practical AI if you build AI systems, and add Dwarkesh Podcast for the week's one interview worth blocking time for. Skip anything still recommending Hard Fork: it stopped publishing in September 2026.
Dupple X is the newsletter behind this kind of tracking, or start the free trial. No podcast paid for a slot on this page, and the about page names the editorial team.
Cite this: Dupple, "9 Best AI Podcasts in 2026 (Verified Still Publishing)", September 2026.
