AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Why I Built A Second Brain For My AI Agents on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get the latest gadgets delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

Thorsten Meyer says he built a shared Second Brain so his AI agents can access and update the same long-term memory across sessions. As of November 21, 2026, he reports 1,278 pages and a 97.4% top-one recall score; those are figures from his own system, not an independent evaluation.

Thorsten Meyer has published details of a shared memory system for his AI agents, built from markdown files, a searchable index and Git version history. He says the system held 1,278 pages and recorded a 97.4% top-one recall score as of November 21, 2026, offering a detailed account of one individual’s approach to helping agents retain information between sessions.

Meyer describes the system, which he calls his Second Brain, as a knowledge base that both he and his agents can read and update. Each memory is stored in a plain markdown file, with a single INDEX.md listing the files and summarizing their contents. The files are kept in Git, which Meyer says preserves a history of changes. He says the setup uses neither a vector database nor a cloud service.

For a recall, a script called recall.py scores the index entries before opening the best-matching page. It then returns the section it considers relevant, along with the source page. Meyer says this lets him check the material behind an agent’s answer. A separate script, remember.py, writes a new page, its index entry and a log entry together, a design intended to keep the catalogue and stored notes in sync.

Meyer reports that the system contained 1,278 pages and 1,761 links on November 21, 2026, and had recorded 1,363 recalls. He says its “top-1” score was 97.4%, meaning, by his description, that the first page selected was the right one in that share of recalls. The article does not provide an independent test or further methodological detail for that figure.

At a glance
reportWhen: Published November 21, 2026
The developmentThorsten Meyer published a description of a personal, file-based memory system shared by the AI agents he runs.

A Checkable Memory Across Agents

The design addresses a practical limitation Meyer identifies: an agent’s working context is tied to what it can access during a session, while information from earlier sessions may need to be supplied again. A shared store can reduce repeated explanations and let multiple agents draw on the same project facts, decisions and preferences, according to his account.

Keeping memories in readable files also changes how users can inspect and maintain them. Meyer’s approach makes it possible for an owner to edit a record directly and use Git history to review changes, rather than relying solely on a tool-specific memory store. Including a source page with recalled information provides a route to check what an agent used. Those are design advantages described by Meyer, not evidence that the system will improve the accuracy of agents generally.

The trade-off is that the arrangement depends on careful organization and is designed around one owner. Meyer acknowledges that it is not built as a team system. Anyone considering a similar setup would need to weigh its portability and inspectability against the work of maintaining files, an index and access rules.

Products Worth Considering

LG UltraGear 34-Inch Curved QHD Gaming Monitor, 160Hz, FreeSync Premium

LG UltraGear 34-Inch Curved QHD Gaming Monitor, 160Hz, FreeSync Premium

As an affiliate, we earn on qualifying purchases.

From Session Context to Files

Meyer frames the project as a response to agents that begin sessions without knowing what happened in earlier work. He says existing memory features can be difficult to inspect or edit, and may be separate for each tool. His alternative is to keep one shared collection that he can read directly and that the agents he runs can consult.

The system’s basic workflow is deliberately small: check the index, open a relevant page, and return the useful section with its source. New information enters through a single write process that updates the page, index and log together. Meyer also describes nine views for monitoring and organizing the material, including a daily brief, a triage process for session notes, a history of changes and a live view of agent activity. These are interfaces around the underlying files, rather than a replacement for them.

“I got tired of being their memory. So I built them one.”

— Thorsten Meyer, describing the problem in his November 21, 2026 article

Limits of the Reported Score

The 97.4% top-one score is a figure Meyer reports for his own system. His article does not specify the number or composition of questions used to calculate it, how a correct page was judged, or whether the score was measured across a fixed test set. It should not be read as an independently verified benchmark or a general measure of agent accuracy.

The account also does not establish how the system performs with a larger group of users, different agents or a much larger knowledge base. Meyer says the design is built for one owner, and does not describe team permissions, conflict resolution between users, or broader security testing. The reported page and link totals are a snapshot dated November 21, 2026; the article does not say how quickly the store is growing.

How Meyer Says It Is Used

Meyer’s article presents the Second Brain as an operating system for his ongoing work with agents, rather than announcing a commercial product or a formal evaluation. He describes views for reviewing new captures, tracking changes and observing agents as they recall or add material. The next practical test is whether the workflow continues to keep its index useful as the collection and agent activity grow.

For readers adapting the idea, the account points to a basic sequence: store related facts in readable files, maintain a concise index, make recall return sources and keep additions logged. Meyer’s description does not include a public test protocol or a team-ready release, so questions about performance beyond his personal use remain open.

Key Questions

What is Meyer’s Second Brain?

It is a personal knowledge base that Meyer says he and his AI agents share. Memories are kept in markdown files with a central index and Git history.

How does an agent retrieve a memory?

Meyer says recall.py first scores lines in the index, then opens the best-matching page and returns a relevant section with its source.

What does the 97.4% score measure?

Meyer calls it a top-one recall score: the share of recalls for which the first page selected was, by his measure, the right one. His article does not give enough testing detail to independently assess the result.

Is the system designed for teams?

No. Meyer says the setup is designed for one owner. His article does not describe shared permissions or processes for multiple users managing the same memory.

Does the system use a vector database or cloud service?

Meyer says it uses plain markdown files and Git, with a script that searches the index. He says the system uses neither a vector database nor a cloud service.

Source: ThorstenMeyerAI.com

You May Also Like

Build a Second Brain: The Basics

Unlock the essentials of building a second brain to organize your knowledge effectively—discover how to transform chaos into clarity today.

Digital Warranty Tracking That You’ll Actually Use

A practical digital warranty tracking system can transform your process, but discovering the right features makes all the difference in…

Cloud‑Syncing Large Video Projects Without Bandwidth Caps

Harness the right strategies to cloud-sync large video projects without bandwidth limits—discover how to keep your workflow seamless and efficient.