This has been running for a long time by podcast standards, with a back catalogue in the high hundreds of episodes, and it has outlasted a great many shows that launched with more attention. The current host, Jon Krohn, has a doctorate, has taught extensively and has written on deep learning, and that background shows in the questions. He can follow a technical answer and ask the next question rather than nodding and moving to the next item on his list, which sounds like a low bar and eliminates most of the competition. The guest mix is the show's strength.
Rather than an unbroken run of founders explaining why their company is important, you get research scientists, working practitioners, academics, tool builders and people who have spent a career in a specific corner of applied statistics. That variety means the show covers what data science actually looks like as a job, which includes an unglamorous amount of data cleaning, stakeholder management and models that did not work, alongside the interesting parts. The two format structure is a good design. Longer interview episodes run through a guest's work in depth, and shorter episodes cover a single topic without a guest.
The short ones are underrated. A focused twenty minutes on one concept, delivered by someone who teaches for a living, is often more useful than two hours of conversation, and they are easy to fit into a commute. The show notes deserve credit. Links to papers, tools, books and the guest's work, maintained consistently rather than sporadically.
For a show that mentions a lot of resources this is the difference between a podcast you listen to and one you learn from. The technical range is reasonably wide. Classical statistics and experimental design, machine learning engineering, deep learning, large language models, the operational side, and a fair amount on the career and organisational realities of the field. That breadth suits someone who wants to stay generally current rather than to go deep on one thing.
Now the criticisms, and the first is the most consistent complaint anyone has about this show. There is a lot of promotion. The host's own courses, books and training material come up frequently, and while it is usually within an episode rather than dressed up as content, the cumulative effect over a few episodes is noticeable. This is a commercial operation attached to a training business, and that shapes the show.
It is not deceptive and it is present. Interview podcasts live and die by their guests and this one has the usual variance. Some guests are excellent communicators with something specific to say. Others give the same set of prepared answers they have given elsewhere, and forty minutes goes by without a concrete claim.
There is no way around this in the format and it does mean you should choose episodes rather than subscribing and listening to everything. The balance tilts towards career and industry conversation more than the title suggests. Discussions about breaking into the field, hiring, what employers want, how the market is changing. That is useful material and it is not technical material, and someone tuning in for depth on a method will sit through a fair amount of it first.
Some episodes are effectively vendor appearances. A guest from a company that sells a tool, discussing the problem their tool solves. These are usually informative in a limited way and they follow a predictable shape and you learn to recognise them from the guest's job title. And the breadth cuts both ways.
Covering everything means going deep on nothing. Someone who works specifically on retrieval systems, or on causal inference, or on any well defined specialism will find the coverage of their area introductory. This is a generalist show for a generalist audience and it is honest about that. My three point seven is for a well hosted, consistently produced show with better guests and better preparation than most of the field, held back by commercial promotion that is always in the room, guest variance the format cannot fix, and a breadth that keeps it from ever going deep.
Worth having in a rotation. Not worth listening to completely.