Best AI and Automation Podcasts in 2026
4 min read · Updated Aug 14, 2026
Podcast listening keeps growing: Edison Research's Infinite Dial 2025 study found 55% of listeners aged 12 and older now consume a podcast monthly, more than double the 24% recorded in 2017. The AI and automation space has enough shows now to skip the marketing-heavy ones and go straight to people building in production. Below are six worth a slot in a commute or a long build queue, sorted by what each is actually good for.
Key takeaways
- Interview-heavy shows are fine at 1.25x–1.5x speed, drop to 1x for technical deep-dive episodes where the pacing already moves fast.
- Most "best AI podcast" lists are recycled from a single old post, check the last few episode dates before subscribing to anything.
- Different shows serve different needs: research depth, applied engineering, industry direction, or fast-moving news.
- A show that was great a year ago can quietly decline, re-evaluate your queue every few months instead of subscribing once and forgetting.
- Video versions of the same show rarely add anything for a commute or a build-queue listen, audio is fine unless you specifically want to watch a screen share.
Quick picks
| Podcast | Focus | Format | Best for |
|---|---|---|---|
| Latent Space | AI engineering and LLM application building | interview, technical deep dives | developers shipping LLM products who want the detail other shows skip |
| Practical AI | applied AI and ML for working engineers | weekly conversational | broad practical coverage without heavy hype |
| The TWIML AI Podcast | ML research and industry interviews | long-form interview | going deeper on the research side of things |
| No Priors | AI industry and startup perspective | interview | the business and startup angle on where AI is headed |
| The Cognitive Revolution | fast-moving AI news and analysis | frequent short-to-medium episodes | keeping up week to week without falling behind |
| Software Engineering Daily | broad software engineering, regular AI and automation episodes | daily interview | automation and infra topics alongside AI |
What makes a good AI podcast in 2026
The field moves fast enough that a show recorded eight months ago can already sound dated, so recency matters more here than in most podcast categories. Look for hosts who have actually shipped something (a product, a paper, an open-source library) rather than hosts who mainly interview other podcasters about the field in the abstract.
The opinion I will defend
A story: clearing the backlog during a long batch job
During a backfill job in late 2024, reprocessing a few years of historical support tickets through a fine-tuned classifier that took about six hours to finish, I worked through a backlog of nineteen unlistened podcast episodes waiting for it to complete. About half of them were shows I had been keeping in my queue purely out of habit. I unsubscribed from three of them that same afternoon and never missed them.
Frequently asked questions
Frequently asked questions
Do I need to watch these podcasts on video, or is audio enough?
Audio is enough for almost all of these, they are conversation-driven, not screen-share-driven. Video adds little unless a specific episode is walking through a live demo.
How often do these shows release new episodes?
It varies by show, from daily to weekly to a slower cadence for the most research-heavy interviews. Check the feed directly since release schedules shift over a show's life.
Are these podcasts beginner-friendly?
Practical AI and Software Engineering Daily are the most approachable for someone newer to the field. Latent Space and the research-heavy shows assume more background.
Should I subscribe to all of these at once?
Start with two that match your current focus (engineering versus industry direction), and add more only once you are actually keeping up, an overloaded queue is how good shows get abandoned unheard.
What listening speed do you actually recommend?
1.25x-1.5x for interview-heavy episodes, 1x for dense technical deep dives where speeding up costs comprehension rather than just time.