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

🔍 Read the full analysis: How To Help AI Agents Remember With A Second Brain on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get audio and creator gear 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 retrieve and add long-term memories across sessions. The system uses Markdown files, Git and a one-line-per-page index; Meyer reports a 97.4% top-1 recall score as of Nov. 21, 2026. The account describes one person’s setup, not an independently tested product or a team-ready service.

Thorsten Meyer has described a shared “Second Brain” that lets the AI agents he runs retrieve and add information across sessions, addressing the loss of context when a session ends. In an article published Nov. 21, 2026, Meyer said the system uses Markdown files, Git and a searchable index, and reported a 97.4% top-1 recall score. The figures and performance measure are his own account, not an independent evaluation.

Meyer’s design stores each cluster of facts—such as information about a project or a decision—in a plain-text Markdown page kept in Git. A file named INDEX.md lists each memory on a single line. Meyer says agents check those index entries first, then a script called recall.py scores them without opening the pages. It opens the highest-scoring page and returns the section that appears to answer the question, along with the source page.

New information is added through remember.py, which Meyer says writes a page, its index entry and a log entry together. The aim is to keep the catalogue aligned with the pages and preserve a record of changes. Git also retains the files’ history. Meyer says the system does not use a vector database or cloud service; he presents it as a folder of text files that can be opened and checked with ordinary editing tools.

As of Nov. 21, Meyer reported 1,278 pages, 1,761 links and 1,363 recalls. He said the system’s “top-1” measure—how often its first selected page is the right one—was 97.4%. He also described nine views for managing or observing the system, including a daily brief, a capture-triage process, a change log, a graph of linked pages and a console for monitoring services and machines.

At a glance
reportWhen: Published Nov. 21, 2026; figures report…
The developmentThorsten Meyer published a description of a file-based memory system shared by his AI agents, reporting its design and usage figures as of Nov. 21, 2026.

A Shared Memory Across Sessions

The approach addresses a practical limitation of AI-agent workflows: information available inside one session may not be available when another begins. A shared store could reduce repeated explanations and help agents apply prior decisions or project details, while making it possible for the owner to inspect the source behind a recalled answer.

Meyer’s account also highlights a trade-off between transparency and upkeep. Plain files are readable, editable and stored with a change history, but the system depends on keeping pages and index entries useful and current. Meyer describes the setup as designed for one owner rather than a team. It does not establish that the same workflow will work at larger scale, or that its reported recall rate would hold across other users, agents or questions.

Products Worth Considering

Sony WH-1000XM6 Wireless Noise-Cancelling Headphones – Midnight Blue

Sony WH-1000XM6 Wireless Noise-Cancelling Headphones – Midnight Blue

As an affiliate, we earn on qualifying purchases.

From Session Context to Stored Notes

AI agents typically work with information available in their active context window. When that session ends, its contents may no longer be available to a later session unless they have been saved elsewhere. Meyer says existing memory features did not meet his preference for a single store he could read, edit and audit across the agents he uses.

His response is not a new commercial product announcement in the supplied account, but a description of a personal workflow and its components. The method is index first, files second: rather than opening every note, the script uses brief index entries to choose a page, then returns a relevant section and its source. Meyer also describes a triage step for deciding whether information captured from sessions should become a lasting memory, be merged into an existing page or be discarded.

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

— Thorsten Meyer, in his Nov. 21, 2026 article

How the Reported Score Was Measured

The account does not provide an independent test of the 97.4% top-1 score, the questions used to calculate it, or the evaluation period and comparison baseline. Meyer defines the measure as how often the first page selected is the right one, but does not explain how “right” is judged or whether results include different kinds of queries.

It is also unclear how the system handles conflicting or outdated memories, access controls, or concurrent changes when multiple agents write to the store. Meyer says the setup is built for one owner, and the supplied description does not establish how it would perform in a shared workplace or with a larger number of agents. The figures are reported for Nov. 21, 2026, and may change as the system is used.

Testing the Workflow Beyond One Owner

Meyer’s article describes the system and points readers to an explainer video showing its views and workflows. The next practical step for readers interested in the method is to review how its index, recall script and writing process fit their own work before relying on it for important information. The account does not announce a release date, public product, independent benchmark or planned team version.

For any wider assessment, further details would be needed on the scoring method, the handling of incorrect or sensitive memories, and whether the approach remains reliable as the collection grows. Until then, the reported page count and recall score describe Meyer’s system at a specific point in time rather than a general performance guarantee.

Key Questions

What is the Second Brain?

It is Meyer’s name for a shared long-term memory store that his AI agents can search and add to across sessions. He describes it as Markdown pages, an index and scripts for recalling and writing information.

Does it use a vector database or cloud service?

Meyer says the system uses plain Markdown files kept in Git, with no vector database or cloud service in the setup he describes.

What does the 97.4% figure mean?

Meyer defines top-1 recall as how often the first page selected by the system is the right one. The article reports 97.4% as of Nov. 21, 2026, but does not provide an independent benchmark or full testing methodology.

Can a team use the same setup?

The article does not establish that. Meyer says his current arrangement is built for one owner, not a team, and gives no details about team permissions or coordinated editing.

Source: ThorstenMeyerAI.com

You May Also Like

How OpenAI, Google, Microsoft And Anthropic’s Opt-Out Works

OpenAI, Google, Microsoft and Anthropic plan to recognize one robots.txt opt-out for AI training, but coverage and start dates differ.

Week Three — Foundation model vs Brownian motion. Kronos on five-minute BTC.

Kronos foundation model tested against Brownian baseline for 5-minute BTC predictions; results show no significant outperformance in recent trading simulation.

Seed AI Chief Zhou Chang Reportedly Sets Out On His Own; ByteDance Denies It

ByteDance has denied a report that Seed AI research leader Zhou Chang left to start a company. His current role and the report’s evidence remain unclear.

Modern Scaling Laws: From Chinchilla to Efficiency Frontiers

Keen insights into modern scaling laws reveal how size and data strategies push AI efficiency frontiers, compelling you to explore further.