Six practitioner-led AI/ML newsletters worth a subscription, chosen because they’re written by people who actually build, not by content teams repackaging Twitter. Each entry below has the author, the rough cadence, and an honest note on what you’ll get out of it. None are affiliate-tagged. None will sell you a course in every issue.
Why these six ML newsletters
Most “best AI newsletters” lists rank by reach, not by signal. This one ranks by signal. The filter: a single named author or small team, a clear beat, technical depth instead of news aggregation, and a cadence that doesn’t try to manufacture urgency. Six entries made the cut; another sixty didn’t.
If you’re already subscribed to one of these and considering another, this list helps you decide which adds something new rather than more of the same. Authors and cadences are noted explicitly so you can avoid duplicating coverage.
1. Gradient Ascent by Art of Saience
A practitioner AI newsletter with a clean beat: applied techniques, explained in depth. Less news-cycle-driven than most; more “here’s how this technique actually works” than “here’s what launched this week.” The writing assumes you build, not just read.
Subscribe if you want a calmer cadence and a deeper-than-average technical treatment per issue. The archive is browsable without subscribing, so read a few issues first to confirm the fit.
Why it matters: the antidote to “10 things that happened in AI this week.” Author: Art of Saience team. Cadence: weekly.
↔ Subscribe to Gradient Ascent
2. DecodingML by Paul Iusztin
Paul Iusztin’s newsletter on building ML and LLM systems, data pipelines, model architecture, the engineering around shipping. Co-author of the LLM Engineer’s Handbook (O’Reilly, entry #3 in our Books roundup); the newsletter is the ongoing version of the same engineering practice.
If “ML engineering in production” is your beat, this is the one to subscribe to. The technical depth is real; the writing is clear; the topics reflect what an actual practitioner is currently working on.
Why it matters: the ML engineering newsletter, written by an author with a book in the field. Author: Paul Iusztin. Cadence: weekly-ish.
3. Deep (Learning) Focus by Cameron Wolfe
Cameron Wolfe’s long-form newsletter on deep learning research, papers, architectures, the conceptual history of how we got here. Each issue is a deep dive rather than a list. If most AI newsletters feel shallow, this one is the corrective.
Subscribe if you want to actually understand the research literature instead of skimming abstracts. Not the right fit if you want news-cycle coverage; perfect if you want depth.
Why it matters: the deep-research newsletter for people who want to read papers, written by someone who reads them carefully. Author: Cameron Wolfe. Cadence: weekly long-form.
↔ Subscribe to Deep (Learning) Focus
4. NeoSage by Shivani
Shivani’s newsletter on applied AI and the practitioner side of the ecosystem. Strong on tools, workflows, and the lived experience of building AI products. Less research-flavored than Cameron Wolfe’s newsletter; more engineering-flavored.
Subscribe if you want a complementary voice to the research-heavy newsletters, one that focuses on the daily reality of shipping AI rather than the academic literature.
Why it matters: the practitioner-side complement to the research newsletters. Author: Shivani. Cadence: weekly.
5. Jam with AI by Shirin & Shantanu
A two-author newsletter covering the applied AI ecosystem with a slightly more playful voice than the others on this list. Useful for staying current on what’s shipping without the breathless hype tone that dominates most AI coverage.
Subscribe if you want one newsletter that’s a bit more accessible than DecodingML or Deep (Learning) Focus but still written by people who build. Pairs well with the heavier technical newsletters above.
Why it matters: the approachable applied-AI newsletter. Authors: Shirin & Shantanu. Cadence: weekly.
6. Data Hustle by Sai
Sai’s newsletter on data engineering and the data side of AI systems, the layer underneath most LLM applications that most LLM coverage ignores: pipelines, storage, retrieval infrastructure, the parts that make or break production.
Subscribe if you’ve realized that the bottleneck on most LLM applications isn’t the model, it’s the data engineering. This is the newsletter for that layer.
Why it matters: the data-engineering layer that most AI newsletters skip. Author: Sai. Cadence: weekly.
How to subscribe
Six newsletters is too many to subscribe to if you’re trying to reduce inbox load. The honest approach: pick two. One research-flavored (Cameron Wolfe’s Deep (Learning) Focus) and one engineering-flavored (DecodingML or Data Hustle). That covers the two ends of the practitioner spectrum without duplicating coverage. Add a third (Gradient Ascent or NeoSage) if you want broader applied coverage; add Jam with AI if you want a more accessible voice in the rotation.
All six are free; most offer paid tiers for archive access or community. The free tiers are sufficient for staying current. None of the links above are affiliate-tagged; we don’t take referral fees from any of these authors.
Six newsletters, two archetypes. Research-deep on one end, engineering-practical on the other. Pick one of each; skip the rest. Your inbox will thank you.