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

🔍 Read the full analysis: My AI Agents Share A Second Brain—Here’s How on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get the little things that make your day 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 describes a personal knowledge base that lets his AI agents share long-term memory through Markdown files and a searchable index. As of Nov. 21, 2026, he reports 1,278 pages and a 97.4% top-1 recall score; the system is designed for one owner, and its performance has not been independently verified.

Thorsten Meyer has described a shared memory system that lets the AI agents he uses consult and update the same personal knowledge base, rather than starting each session without information from earlier work. In a post published Nov. 21, 2026, Meyer said the system contains 1,278 pages and has a reported 97.4% top-1 recall score, figures he presents as measures of its size and retrieval performance.

Meyer’s system, which he calls his Second Brain, stores knowledge in ordinary Markdown files tracked with Git. Each page groups related information, such as facts about a project or a decision. A single INDEX.md file lists each memory in one line, allowing agents to locate relevant pages without searching every file.

According to Meyer, a script called recall.py scores the index entries, opens the highest-scoring page and returns the section relevant to an agent’s question, along with its source. A separate script, remember.py, adds a memory, its index entry and a log entry together. Meyer says that process is intended to prevent the index and the underlying notes from falling out of sync.

The system also includes nine views for monitoring and managing the knowledge base. Meyer’s list includes a daily brief, a triage process for deciding what to retain from sessions, a history of changes, a graph of links between pages and a console for monitoring services. He says the views are built around the same files and index, rather than separate memory stores.

As of Nov. 21, Meyer reports 1,278 pages, 1,761 links and 1,363 recalls. He defines the 97.4% top-1 figure as how often the first page selected by the system is the right one, according to the system’s own measure. The material does not provide a testing method or independent evaluation for that score.

At a glance
reportWhen: Published Nov. 21, 2026
The developmentThorsten Meyer published a description of a shared, file-based memory system for the AI agents he uses.

One Memory Across Multiple Agents

A shared knowledge base can reduce the need to repeat project details, prior decisions and terminology every time an agent session begins. It also gives the user a way to inspect the material behind an answer: Meyer says each recall identifies its source page, which can be opened and checked.

The design prioritizes readability and portability. Plain text files can be viewed or edited with common tools, while Git records changes over time. That differs from systems whose stored memories are difficult for users to inspect or move, although the article does not compare Meyer’s setup with particular commercial products.

Meyer also identifies a limit: the arrangement is built for one owner, not a team. A shared repository may make information available to multiple agents, but it does not by itself settle questions about permissions, conflicting edits, privacy or how a group should approve and maintain shared knowledge.

Products Worth Considering

Levoit Vital 200S-P Smart Air Purifier for Large Rooms (Up to 1875 sq ft)

Levoit Vital 200S-P Smart Air Purifier for Large Rooms (Up to 1875 sq ft)

As an affiliate, we earn on qualifying purchases.

From Session Context to Stored Notes

AI agents generally work with information available in their active session. Meyer’s account begins with the practical problem that, after a session ends, an agent may not carry forward the user’s earlier decisions or working conventions. He says he created the Second Brain because he uses many agents and did not want to repeatedly provide the same background.

The approach is deliberately limited to Markdown files, an index and scripts. Meyer says it uses neither a vector database nor a cloud service. Git provides a history of edits, while the index serves as the first step in retrieval. The article presents this as his own working system and explains its design; it does not report a controlled study or a comparison against other memory architectures.

Meyer’s post also references an explainer video and screenshots of the system’s views. It says personal information in the video and frames is protected, and that some names and contents are masked or replaced with cover names.

“One memory, shared by every agent I run, that I can open and read myself.”

— Thorsten Meyer, describing the system in his Nov. 21, 2026, post

How Reliable Is the Recall Score?

The reported 97.4% top-1 score is Meyer’s own measure. The post does not explain the size or composition of the test set, how a correct page was determined, or whether the score was evaluated independently. It should not be read as a general measure of accuracy for AI answers; it describes the system’s reported ability to select a relevant page first.

The article also does not specify how the setup handles simultaneous edits, access controls or recovery if a file or index entry is damaged. Meyer describes a workflow intended to keep entries synchronized, but provides no independent audit of the system’s operation or security. The reported totals are a snapshot as of Nov. 21, 2026, not a continuing performance guarantee.

What Meyer’s Walkthrough Shows

Meyer’s post points readers to an explainer video showing the system’s nine views, including its map, daily brief, live activity and control desk. That walkthrough is the next available source of detail on how he uses the setup in practice. The post does not announce a product release, public service or planned team version.

For readers considering a similar approach, the described workflow suggests a practical next step: decide what information should persist, organize it into readable pages, and maintain a concise index that agents can search. Any claims about usefulness or accuracy would need to be tested against the reader’s own tasks; Meyer’s reported figures apply to his system and are not independently verified in the source material.

Key Questions

What is Meyer’s Second Brain?

It is a personal knowledge base that Meyer says he and his AI agents share. It stores information in Markdown pages, with an index to help agents find relevant material.

How do the agents retrieve information?

Meyer says recall.py scores lines in the index, opens the best-matching page and returns a relevant section with its source. The agent does not search every page at the first step.

What does the 97.4% figure measure?

Meyer describes it as the system’s top-1 recall score: how often the first page it selects is the right one, by the system’s own measure. The post does not give a test protocol or independent verification, so the figure should be treated as a reported result.

Is the system built for teams?

No. Meyer says the setup is designed for one owner, not a team. The post does not describe team permissions, shared review or rules for resolving conflicting edits.

Does the setup use a cloud service or vector database?

Meyer says it uses plain Markdown files tracked in Git, with a single index and scripts, and does not use a vector database or cloud service.

Source: ThorstenMeyerAI.com

You May Also Like

5 food trends driven by lifestyle shifts around the world – FoodNavigator-Asia.com

Five key food trends are emerging worldwide, driven by shifts in consumer lifestyles, according to FoodNavigator-Asia. These trends impact food innovation and market strategies.

Hexo Labs Builds Self-Improving AI That Learns On The Fly

Hexo Labs has announced the creation of an AI system that can learn and adapt on the fly, marking a significant step in autonomous machine learning technology.

Why Your Approval Doesn’t Let An AI Agent Book A United Flight

United’s revised contract bars unauthorized third-party AI agents from booking or managing tickets, even when a passenger supervises or approves.

Connor Murphy

Recent searches and social media activity have brought Connor Murphy back into public focus amid ongoing speculation about his well-being.