MemPalace Organizes AI Memory With a Palace Metaphor

Matthew Berman introduces an open-source memory project and its reported retrieval score, while noting that AI memory tools already exist.

MBMatthew BermanSpeaker · Forward Future
Channel
Matthew Berman
Published
Apr 8, 2026
Video
1:13
Read
1 min
Slides
1
Original
English
Open on YouTube
Scroll to read and the video shrinks to the bottom-right corner. Captions follow your reading language.
Reading languageAvailable in 2 languages
You're reading the original English.
SUMMARY

In 30 seconds

Matthew Berman introduces MemPalace, a project associated with Milla Jovovich that organizes AI memories using a palace metaphor. He cites a 96.6% LongMemEval result and calls it the best memory system. The project describes that figure as raw Recall@5 for its retrieval setup; it is not a general proof of best end-to-end memory performance.

  1. 1An open projectBerman introduces MemPalace and associates it with Milla Jovovich. He describes the project as free and open source.
  2. 2Remembering across sessionsHe frames the project around information lost between AI conversations, while acknowledging that other memory systems already address the problem.
  3. 3The palace metaphorThe project groups memories into wings, halls and rooms. Its reported benchmark figure concerns a particular retrieval metric.

Takeaways

  1. MemPalace uses wings, halls and rooms as labels for organizing memories.
  2. Berman notes that AI memory systems already exist, even though he considers them imperfect.
  3. The project's reported 96.6% is a raw Recall@5 figure on a specific LongMemEval retrieval setup.
  4. A retrieval score and a project ranking claim do not establish superiority across all memory tasks or configurations.
FULL TEXT

Read the edited text by chapter

Transcript only
01

Introducing MemPalace

Berman introduces the project and its association with Milla Jovovich. He frames persistent memory as a problem for AI agents.

Matthew Berman introduces MemPalace, an open-source AI memory project associated with Milla Jovovich. He calls it the best agent memory system and cites a reported 96.6% LongMemEval result. The project describes that figure as raw Recall@5 under a specific retrieval setup; it does not establish superiority for every end-to-end memory task.

Berman frames the problem as conversations, decisions and debugging context being lost when an AI session ends. He acknowledges that memory systems already exist, though he finds them imperfect. MemPalace is presented as another way to organize and retrieve past material.

02

Organization and benchmark claim

He explains the palace metaphor and cites a 96.6% benchmark figure. That number is a specific retrieval metric, not a universal ranking.

The memory-palace structure
SLIDE 1 / 1The memory-palace structureThe project's description groups memories into wings, halls and rooms.

The project borrows the memory-palace metaphor: wings represent people and projects, halls represent types of memory, and rooms represent specific ideas. Berman compares it to arranging ideas in an imagined building. This is an organizational metaphor, not proof that the software reproduces how human memory works.

More from this channel

12 slides34:21EN
Matthew Berman· Oct 27, 2026Matthew Berman Examines Anthropic’s IPO Filing, AI Growth and GovernanceMatthew Berman, Sam Altman · Forward FutureThe presenter reviews the S-1’s financial metrics, compute strategy and voting structure while questioning the tension between rapid AI development and safety.AnthropicAI SafetyEnterprise AI+2
0:26EN
Matthew Berman· Oct 25, 2026Sam Altman on Human Values and Problem-Solving in an AI WorldSam Altman · OpenAIIn a short interview excerpt, Altman considers the enjoyment of solving hard problems and the human role in interpreting results.Human-AI CollaborationFuture of WorkAI Strategy
2 slides0:36EN
Matthew Berman· Oct 25, 2026Kimi K3.1 and a Rubik’s Cube Solver DemonstrationMatthew Berman · Forward FutureMatthew Berman shows a scrambled virtual cube being solved and gives the Kimi K3.1 test a favorable assessment.EvalsCoding agentsOpen-weight models
2 slides0:21EN
Matthew Berman· Oct 24, 2026An AI-Generated Kimi Benchmark Chart Used Invented ScoresMatthew Berman · Forward FutureMatthew Berman asks for a graph of Kimi K3.1 Pro benchmarks, then discovers that the displayed figures are illustrative placeholders.BenchmarksEvalsGenerative media
6 slides25:50EN
Matthew Berman· Oct 24, 2026Kimi K3.1 Pro Tested: Strong 3D Demos, Uneven Results and Competitive PricingMatthew Berman · Forward FutureMatthew Berman tests the open-weight model on coding, 3D rendering and visual explanations, then compares benchmark results and prices.Open-weight modelsBenchmarksAI cost & performance+1
2 slides32:20EN
Matthew Berman· Oct 23, 2026Who Is Local AI For? The Costs, Trade-offs and a Hybrid FutureMatthew Berman · Forward FutureMatthew Berman weighs data control and independence against hardware cost, setup complexity and inference speed.Local inferenceOpen-weight modelsAI cost & performance+1