Could A Second Brain Help AI Agents Stay In Sync?
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Could A Second Brain Help AI Agents Stay In Sync? on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get smart everyday buys 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 personal knowledge base that lets his AI agents share long-term memory across sessions. As of Nov. 21, 2026, it held 1,278 pages; Meyer reports a 97.4% top-one recall score, but the system is designed for one owner and its performance has not been independently verified.

Thorsten Meyer has described a shared memory system for his AI agents, designed to carry information such as project details and past decisions between sessions. In an article published Nov. 21, 2026, Meyer said the system contained 1,278 pages and that its own top-one recall measure was 97.4%; those figures are his, and the source does not provide an independent evaluation.

Meyer calls the system a Second Brain: a personal knowledge base that he and his agents can both read and update. Its records are plain Markdown files stored in Git, with one index file, INDEX.md, listing each memory. Meyer says the design uses no vector database or cloud service, and that the files can be opened and edited with ordinary tools.

When an agent needs information, a script called recall.py scores lines in the index, opens the highest-ranked page and returns the relevant section with its source. Meyer says this approach lets him check which page supports an answer. A second script, remember.py, writes a new page, its index entry and a log entry together, a process he says is intended to prevent the catalogue from falling out of step with the files.

As of Nov. 21, Meyer reported 1,278 pages, 1,761 links and 1,363 recalls. He put the system’s top-one recall score at 97.4%, defining that measure as how often the first page selected is the right one. The article does not specify the test method, sample, evaluation period or comparison baseline for that score.

At a glance
reportWhen: Published Nov. 21, 2026; system figures…
The developmentThorsten Meyer published a description of a shared, file-based memory system intended to help his AI agents retain and use information across sessions.

A Shared Memory Across Agent Sessions

The system addresses a practical limitation of session-based agents: information available in one conversation may not carry into the next. A shared store could let multiple agents consult the same project facts and decisions, reducing repeated instructions and making it easier for a user to correct a record in one place.

Meyer’s approach also puts readability and traceability ahead of using a specialized hosted memory service. Plain files and Git history allow the owner to inspect changes and, in principle, move or edit the data with familiar tools. That may appeal to users who want greater control over what an agent remembers and the basis for its answers.

There are trade-offs. Meyer says the setup requires discipline and is built for one owner, not a team. Its usefulness for other users, larger groups or different workloads is not established by the account. The reported recall score is a result from Meyer’s own system, not evidence that shared file-based memory will perform similarly in other settings.

Products Worth Considering

Vari ComfortEdge 48x30 Inch Adjustable Electric Standing Desk

Vari ComfortEdge 48×30 Inch Adjustable Electric Standing Desk

Apple Mac mini with M2 chip, 16GB unified memory, 256GB SSD, Silver

Apple Mac mini with M2 chip, 16GB unified memory, 256GB SSD, Silver

As an affiliate, we earn on qualifying purchases.

How the Memory System Works

Meyer describes the problem as a gap between an agent’s current context window and long-term memory. When a session ends, information in that immediate working context may no longer be available. He says existing memory features can be difficult to inspect or edit and may be tied to individual tools; the Second Brain is his alternative, shared across the agents he runs.

The project uses nine views to display or manage the knowledge base. These include a map of connections, a daily brief, a triage process for deciding whether session captures should be kept or merged, and a chronicle of changes. Other views show links between pages, live recall and writing activity, and the services or machines used by the system. Meyer describes these as ways to monitor the files and workflows; they are not, by themselves, independent measures of accuracy.

The key operational choice is to make the index the first step in retrieval and to keep each memory focused on a cluster of related facts. Git records file changes, while the scripts handle retrieval and entry. Meyer presents the combination as a way to keep memory inspectable and portable, rather than as a benchmarked replacement for other agent-memory architectures.

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

— Thorsten Meyer

The Limits of the Reported Score

The article does not explain how the 97.4% top-one score was calculated, how many questions or test cases were used, or who judged whether a selected page was correct. It also does not give a time window or a baseline for the 1,363 recalls, so that figure should be read as a reported count rather than a rate or growth measure.

There is no outside assessment in the source material, and no comparative results against other memory systems. It remains unclear how the setup performs as the collection grows, how it handles conflicting or outdated facts, or whether it can support multiple people with different access rights. Meyer’s article describes a personal implementation; it does not establish broader reliability or security outcomes.

Testing the Design Beyond One Owner

Meyer’s article does not announce a product launch, public release or independent test. It presents the system as a working setup he uses with his own agents and explains its files, scripts and views. The next evidence readers would need to judge its wider value would include a documented evaluation method for recall accuracy, results across varied queries, and detail on how memory updates and mistakes are handled.

Any move from a personal knowledge base to a team system would also require answers about collaboration, permissions, privacy and maintenance. Until those details or outside evaluations are available, the reported page count and recall score describe Meyer’s system as of Nov. 21, 2026, not a general performance guarantee.

Key Questions

What is Thorsten Meyer’s Second Brain?

It is a personal knowledge base that Meyer says he and his AI agents share. It stores memories in Markdown files and uses an index to help agents find relevant pages.

How do the agents retrieve information?

According to Meyer, recall.py scores the index entries first, then opens the best-matching page and returns the relevant section with its source. The described process does not search every page before choosing a result.

What does the 97.4% figure mean?

Meyer defines it as the system’s top-one recall score: how often the first page it selects is the right one. The article does not give the test method, sample size or independent verification, so the figure should be treated as Meyer’s reported measure.

Is the system designed for teams?

No. Meyer says the setup is built for one owner, not a team. The article does not describe team permissions or collaborative workflows.

Can readers use or download the system?

The source describes the files and scripts but does not announce a public release, download or product. Whether the implementation is available to others is not stated.

Source: ThorstenMeyerAI.com

This content is for general information only and is not financial, tax or legal advice. Consult a qualified professional for decisions about your money.
You May Also Like

Electric Code Calculator

A new mobile and web app is being developed to provide electricians with fast, offline, code-grounded calculations for NEC compliance, addressing industry needs.

Make Halloween Decorating More Creative With AI

AI can generate Halloween visuals and dialogue, but some “AI” props use only sensors and recordings. Here’s what to check before building a display.

The Delegation Ladder: The Four Agentic Loops, and What Each One Lets You Stop Doing

A detailed analysis of the four agentic loops in AI engineering, explaining what each allows you to stop doing and how they shape autonomous AI processes.

8 AI Innovations Changing The Course Of 2026

Eight major AI innovations are transforming industries and society in 2026, with confirmed developments and ongoing advancements shaping the future.