🔍 Read the full analysis: What My AI Agents Gain From A Shared Second Brain on ThorstenMeyerAI.com
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
TL;DR
Thorsten Meyer says he built a shared Second Brain so his AI agents can consult and update the same long-term memory across sessions. The system uses Markdown files, Git and an index rather than a vector database or cloud service; as of Nov. 21, 2026, Meyer reported 1,278 pages and a 97.4% top-one recall score. The account describes one person’s system, not an independently evaluated product or a team-ready service.
Thorsten Meyer says he has built a shared long-term memory for the AI agents he uses, allowing them to consult and add to the same collection of notes instead of starting each session without prior context. In a Nov. 21 account on ThorstenMeyerAI.com, he described a system made from Markdown files, Git and a searchable index, and reported that it contained 1,278 pages with a 97.4% top-one recall score.
Meyer’s system, which he calls a Second Brain, stores information in plain-text files organized around clusters of facts, such as a project or a decision. A single file, INDEX.md, lists each memory in one line. Meyer says the system does not use a vector database or cloud service, and that the files can be opened and edited with ordinary tools. Git records changes over time.
When an agent needs information, a script called recall.py first scores the index entries, then opens the best-matching page and returns the relevant section with its source. Meyer says this lets him check where an answer came from. A separate script, remember.py, writes a new page, its index entry and a log entry together, a design intended to keep the catalogue aligned with the stored notes.
Meyer reported 1,278 pages, 1,761 links and 1,363 recalls as of Nov. 21, 2026. He said the system’s top-one recall score—his measure of how often the first page selected is the right one—was 97.4%. The article does not provide an independent test, evaluation method or comparison benchmark for that figure.
The design addresses a practical limitation of agents that operate within a session context window: information available during one conversation may not carry into the next. A common store could reduce repeated explanations and help different agents use the same project details, terminology and past decisions, according to Meyer’s account.
Keeping the material in readable files also makes the memory inspectable and editable by its owner, rather than tying it to a particular agent’s private memory feature. That may matter to people who want to check the source behind an answer or correct an inaccurate note. These are design advantages described by Meyer, not results from a comparative study.
The approach also has limits. Meyer says it is built for one owner, not a team, and that it requires discipline. A shared store can carry errors or outdated information across agents if its entries are not reviewed. The article does not establish how the system would perform with multiple contributors, much larger collections or different agent platforms.
Products Worth Considering

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

Jackery Explorer 300 Portable Power Station, 292Wh

Bose SoundLink Plus Portable Bluetooth Speaker, Black
As an affiliate, we earn on qualifying purchases.
From Session Context to Persistent Files
Meyer frames the project as a response to agents losing access to information once a session ends. He says existing memory features can be difficult to read, edit or audit, and may differ between tools. His alternative is to put the memory in a folder he controls and allow every agent he runs to use it.
The system’s operation depends on a short index and small scripts, rather than searching every note for each question. Meyer also describes nine views for monitoring the collection, including a map of connections, a daily brief, a triage process for session captures, a change log and a live view of agent activity. Those views help him manage the setup; the source does not say they are part of a general-purpose product available to other users.
“I got tired of being their memory. So I built them one.”
— Thorsten Meyer, writing on ThorstenMeyerAI.com
How the Recall Score Was Measured
The reported 97.4% top-one score is Meyer’s own measure. His article does not specify the test set, how “the right” page was determined, the period covered by the 1,363 recalls, or whether results were independently checked. The number therefore describes the system as Meyer reports it, not a verified benchmark against other memory tools.
It is also unclear how the setup handles conflicts between notes, stale information, permissions or simultaneous edits, and what safeguards prevent sensitive material from being exposed to an agent that should not see it. Meyer says the system is designed for one owner, but gives no results from use by a larger team. The article’s source text ends before completing its discussion of the figures, so further details are not available here.
More Detail Needed for Wider Use
Meyer’s article describes the system and its reported size as of Nov. 21, 2026; it does not announce a launch date, release plan or independent evaluation. Readers considering a similar setup would need to judge how to organize and review records, protect private information, and test whether recall is reliable for their own work.
Further reporting or documentation would be needed to establish how the score was calculated, how the system performs as the collection grows, and whether its scripts and processes can be adapted safely for multiple users. Until then, the reported results should be understood as an account of Meyer’s personal system, rather than evidence that shared agent memory will work equally well in other settings.
Key Questions
What is Thorsten Meyer’s Second Brain?
It is a long-term knowledge base that Meyer says he shares with the AI agents he runs. It stores notes in Markdown files and uses an index to help agents find relevant pages.
How do the agents retrieve information?
Meyer says a script scores entries in the index, opens the best-matching page and returns the section relevant to the question, along with its source.
How large was the system?
Meyer reported 1,278 pages and 1,761 links as of Nov. 21, 2026. He also reported 1,363 recalls, but the article does not state the time period for that count.
Is the 97.4% recall score independently verified?
The source gives no independent verification or detailed testing method. The figure is Meyer’s reported top-one score, meaning his measure of how often the first page selected is the right one.
Can teams use the same approach?
Meyer says his setup is built for one owner, not a team. The article does not describe testing with multiple contributors or explain how permissions and conflicting edits would be handled.
Source: ThorstenMeyerAI.com
