# lennys-newsletterpodcastdata: a markdown transcript corpus you can clone > Lenny's Data is a public repository of podcast transcripts and newsletter posts in plain markdown, with an index.json for titles, dates and word counts. It suits builders testing retrieval or chat products on a real interview corpus, and its ceiling is obvious: the free tier holds 10 posts and 50 transcripts while the paid archive holds hundreds more. **LennysNewsletter/lennys-newsletterpodcastdata** — Public free starter pack for Lenny's Podcast transcripts and Lenny's Newsletter posts, in AI-friendly markdown. - Repository: https://github.com/LennysNewsletter/lennys-newsletterpodcastdata - Website: https://www.lennysdata.com - Stars: 673 · Forks: 177 - Language: Unknown - License: NOASSERTION - Published: 2026-10-21 · Updated: 2026-10-21 · Language: en - Canonical page: https://hysenlabs.com/projects/lennysnewsletter-lennys-newsletterpodcastdata ## A starter pack, not the whole archive The repository publishes a fixed, openly documented slice of a much larger collection. What is included here is 10 newsletter posts and 50 podcast transcripts. The paid archive holds 373 newsletter posts and 322 podcast transcripts, and the difference is stated plainly in the project description: this repo is the public starter pack with a limited number of transcripts and posts. That framing matters more than the numbers do. A starter pack is meant to be enough to build and test against, and the project documentation points at a list of example products built with the data, including a searchable database of skills extracted from the podcast archive, a collection of frameworks and mental models distilled from the episodes, and an AI agent that answers with product advice grounded in the archive. Other examples turn transcripts into visual summaries and a chat interface, or use the interviewing style to generate simulated customer interviews. In other words, the intended user is a builder prototyping one of those shapes. One rule is worth flagging before you plan around it. The documentation states that newsletter posts published within the last 3 months are intentionally excluded from the archive. A corpus that withholds its most recent material is a deliberate choice, presumably to keep the paid tier useful, and it means the public data is always behind the conversation. ## The repo layout is four entries and nothing else The whole repository is four top level items: index.json, newsletters/, podcasts/, LICENSE.md and the README itself. There is no build system, no package manifest, no scripts and no pipeline. Every artifact is plain markdown, which is the reason the project description calls the data AI-friendly and the README says it is easy to use with Claude Code, Cursor and other AI tools. That simplicity is the design decision. A corpus that is markdown on disk can be dropped into a repository, grepped, chunked by a script you wrote, embedded in a vector store or handed to an agent that reads files. Nothing has to be parsed out of a PDF or a proprietary format first, which is where most similar archives lose their usefulness. The cost of that choice is that there is no schema enforcement and no update mechanism visible in the repository. The files are static, so a pipeline over them is a one-time ingest unless you re-clone. The index is the only structured surface, and it is the place to look before writing any code. ## index.json is the only machine readable contract The documentation describes what index.json carries: titles, dates, word counts, newsletter subtitles when available, and podcast guests and descriptions. The word counts are the most useful field for an engineer, because they let you estimate corpus size and filter by episode length without parsing 60 markdown files first. Subtitles are conditional, so any consumer has to treat them as optional, and the same is true of guest metadata on the podcast side. The wording here is careful and worth mirroring. The description says the index includes those fields, and it does not document the exact JSON schema, the key names, the sort order or the encoding. If you write a loader against it, treat it as a best-effort contract: read the file before hard-coding field paths, and write the loader so that a missing subtitle or guest list does not break the ingest. Because the paid archive has the same index structure with more entries, building against the starter index means the loader you write should survive being pointed at the larger set later. Anything that assumes exactly 60 items will not. ## Getting the data and the MCP surface The Quick Start section gives three routes, and only the first is open. You can clone the repository from GitHub with git clone https://github.com/LennysNewsletter/lennys-newsletterpodcastdata.git. If you prefer a ZIP, the documentation says to sign in at lennysdata.com and download it there. The same sign in is required for starter MCP access or for the full archive. That MCP detail is the most interesting thing in the project description beyond the data itself. Free subscribers can connect to a starter MCP at lennysdata.com, and paid subscribers get full MCP access. The point of an MCP server here is that an agent can query the archive as a tool rather than being handed every transcript in context, which is the difference between a demo that works on 50 episodes and one that works on 322. For paid subscribers there is also a private GitHub repo to clone, alongside the archive itself. The project site lennysdata.com is the single front door for all of it, and the upgrade path points at lennysnewsletter.com/subscribe. ## A custom licence and no release history The licence is the sharpest edge in this repository. GitHub classifies it as Other, a custom licence that the platform cannot categorise, and the README gives no summary, only a pointer: see LICENSE.md for usage terms. For a corpus of transcripts and newsletter posts, that is exactly the part you must read yourself, particularly if you plan to publish derived work, redistribute excerpts, or train a model on the text. Nothing in this review substitutes for that file. There is also no GitHub release history for this repository. It is not archived, and the last push was on 2026-10-18, so content is being added or edited, but there is no versioned release stream, no changelog and no tag scheme to pin against. Anyone depending on the corpus in a pipeline has to treat a clone as a moving target or record the commit hash they used. Together these two facts set the boundary of the project. It is a distribution channel with a licence attached, not a maintained library with a release cadence. ## Where this dataset is the wrong starting point Three situations rule it out. The first is production. If you are building a customer-facing product that needs current product management advice, the archive withholds the most recent three months of newsletter posts and the public tier holds only 10 of them, so the material you would query is thin and out of date by design. The second is evaluation. Sixty documents is small enough that a retrieval benchmark built on it risks measuring memorisation rather than retrieval, and small enough that a model with prior exposure to the same episodes can look better than it is. The third is licensing uncertainty. If the terms in LICENSE.md turn out to restrict commercial use of derived content, the archive becomes unusable for the kind of product most people want to build from it, and the paid tier does not remove that question. The real alternative to using this corpus is to build a comparable one from material you already have rights to, or to work directly from the public episode pages. The difference in approach is ownership. This repository gives you a ready-made, pre-cleaned, index-equipped corpus in a format an agent can read today, and it charges you for the breadth and for the licence terms, whereas a self-built corpus costs collection effort but leaves the rights question entirely in your hands. ## Conclusion This repository is worth adopting for prototypes, prompt grounding experiments and evaluation sets, where 10 newsletter posts and 50 transcripts are enough to prove a pipeline and small enough to read end to end. It is not the source for a production product, because the archive you would actually need sits behind a paid subscription, and recent posts are deliberately withheld from it. Before building on it, open LICENSE.md and settle the usage terms yourself, since the licence is a custom one that GitHub cannot classify, and check the index.json word counts to confirm the corpus is large enough for what you intend to measure. ## FAQ ### How many transcripts and newsletter posts are in lennys-newsletterpodcastdata? The free public repository contains 10 newsletter posts and 50 podcast transcripts. The paid archive holds 373 newsletter posts and 322 podcast transcripts, and newsletter posts published within the last 3 months are intentionally excluded. ### What format is the lennys-newsletterpodcastdata dataset in? Everything is plain markdown, split across newsletters/ and podcasts/, with an index.json carrying titles, dates, word counts, newsletter subtitles when available, and podcast guests and descriptions. ### Can I use an MCP server with lennys-newsletterpodcastdata? Yes. Free subscribers can connect to the starter MCP at lennysdata.com, and paid subscribers get full MCP access along with the full archive and a private GitHub repo. ### What licence applies to the Lenny's Data transcripts? It is a custom licence that GitHub cannot classify, and the README gives only a pointer to LICENSE.md for usage terms. Read that file before redistributing excerpts or building derived products. ### Are there GitHub releases for lennys-newsletterpodcastdata? No, the repository has no GitHub releases. It is not archived and the last push was on 2026-10-18, so pin a commit hash if you depend on a specific snapshot. ## Sources - [Issues](https://github.com/LennysNewsletter/lennys-newsletterpodcastdata/issues) - [LennysNewsletter/lennys-newsletterpodcastdata on GitHub](https://github.com/LennysNewsletter/lennys-newsletterpodcastdata) - [Project website](https://www.lennysdata.com) - [README](https://github.com/LennysNewsletter/lennys-newsletterpodcastdata/blob/main/README.md) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/lennysnewsletter-lennys-newsletterpodcastdata