The Compile robot personalizing this week's issue

Compile, tuned to ML and AI

The machine learning newsletter, ranked to your depth

Papers, models, and tools, every week: the ML news that matters to you, at the depth you want, cut to a 15 minute read. Reply to make it deeper, lighter, or narrower. No ads.

First month free. No card to start. Reply to any issue to change it.

What is the best machine learning newsletter?

For understanding, Andrew Ng's The Batch is the gentle weekly standard; for research and policy depth, Import AI; for hands-on technical deep dives, Sebastian Raschka's Ahead of AI. They are all excellent and all broadcast. Compile is the tunable option: it ranks the week in ML against your taste and depth, into a 15 minute read you reshape by replying, so you read one tuned briefing instead of six overlapping ones.

The machine learning newsletters worth reading

The specialists go deep on research or tutorials. One option is built to change to your depth and role.

  • CompileOur pickWeekly · personalized · no adsThe tunable one. Dial ML up or down, ask for papers over product or the reverse, and it follows. One tuned briefing instead of six overlapping subscriptions. First month free, then $1.99.
  • The BatchMost lovedWeekly · DeepLearning.AIAndrew Ng's curated research and news, famous for the Letters from Andrew Ng. Best for understanding, measured and free.
  • Import AIWeekly · by Jack ClarkThe deepest on research, compute, and policy trends, running since 2016. Ends each issue with short AI fiction.
  • Ahead of AIPeriodic · by Sebastian RaschkaLong, tutorial-style deep dives on LLM architectures and fine-tuning. The densest technical pick.

What a personalized issue looks like

For an engineer who ships models and wants applied results over research theory. Ad-free, at your depth.

  • FrontierA clear read on this week's model release, what is genuinely new vs benchmark theaterPast the launch hype: what actually changes for how you build, and where the benchmarks oversell it.
  • TechniquesThe fine-tuning technique quietly beating RLHF on small modelsWhy it works, when to reach for it, and a minimal example. Tuned in because you said you ship models.
  • PapersNew paper roundup: 4 results worth your time, and 2 that are overhypedThe week's research, filtered to what holds up, with the one-line why for each.
  • In productionWhy your RAG pipeline retrieves garbage, and the eval that catches itA concrete retrieval fix and the golden-set test that flags regressions before users do.

Don't like it? Just reply.

The ML specialists send one edition to everyone at their fixed depth. Compile builds yours from a plain-English reply, so it lands at the depth and on the topics you actually want.

Reply to Compile
  • More research papers, less product news.
  • Keep it applied, I ship models.
  • Add a section on fine-tuning and evals.
  • Explain like I am new to ML.
  • Only the frontier model releases.

machine learning newsletter: common questions

What is the best machine learning newsletter?

For understanding, Andrew Ng's The Batch; for research and policy depth, Import AI; for technical tutorials, Ahead of AI. All are strong and all send one edition to everyone. Compile is the tunable alternative that ranks ML to your depth and role, changeable by replying.

Is there a free machine learning newsletter?

Yes, The Batch, Import AI, and TLDR AI are all free. Compile gives you the first month free with no card, then $1.99 a month with no ads, so you get one tuned briefing instead of assembling six free ones.

Daily or weekly machine learning newsletter, which is better?

A daily like TLDR AI is good for headlines but easy to fall behind on. A weekly like The Batch is a digest you finish. Compile is weekly by default, and you can reply to ask for more or less frequency and depth.

Which ML newsletter is best for engineers vs researchers?

Researchers tend to prefer Import AI and Ahead of AI; engineers who just want to keep up lean on The Batch or a daily digest. Compile serves either by tuning: tell it applied over theoretical, or the reverse, and it adapts.

Can I get one AI newsletter instead of six that repeat each other?

That is exactly the point of a personalized one. Instead of subscribing to several broad ML newsletters that cover the same releases, Compile ranks the week to you and sends only what you asked for.

Keep reading

CompileA personalized tech newsletter you steer just by replying. The week in tech and AI, compiled to your taste in 15 minutes. First month free, then $1.99.AI agent newsletterThe best AI agent newsletters in 2026, from AI Agents Weekly to Compile, the one you customize by replying. Agent frameworks, tool-use, and real builds, weekly.Generative AI newsletterThe best generative AI newsletters in 2026, from GenAI Works to MIT Generation AI, plus Compile, the personalized one you tune by replying.Best AI and tech newslettersAn honest 2026 roundup of the best AI and tech newsletters, from TLDR to Morning Brew, and the best personalized paid alternative.

One ML briefing, tuned to you

Start your first month free. Tell Compile your depth and focus, and it becomes the machine learning newsletter only you get.

Start your free monthCompare every AI newsletter