# Wiki Layer

> **TL;DR.** Wiki Layer is an architectural pattern for multi-agent systems: a shared, git-native markdown store that all agents on a team can read from and write to, distinct from each agent's private notebook.

- **Category:** AI / Developer Tools / Agent Architecture
- **Stage:** validating
- **Age:** 71 days
- **Origin date:** 2026-04-06
- **First detected:** 2026-04-25
- **Canonical URL:** https://earlyterms.com/term/wiki-layer
- **Sources:** 6 primary URLs

## Definition

Wiki Layer is an architectural pattern for multi-agent systems: a shared, git-native markdown store that all agents on a team can read from and write to, distinct from each agent's private notebook. It functions as durable team memory — facts and synthesized briefs that persist across sessions rather than evaporating with each context reset.

The term entered public discourse on [April 25, 2026](https://news.ycombinator.com/item?id=47899844) with the [WUPHF](https://github.com/nex-crm/wuphf) Show HN thread, which described shipping "a wiki layer for AI agents that uses markdown + git as the source of truth, with a bleve (BM25) + SQLite index on top." WUPHF hit 124 points and 59 comments within hours, and wikilayer.com had already been registered on April 6.

## Example

In WUPHF's implementation, each agent writes raw task observations to a private agents/{slug}/notebook/ directory, then promotes durable insights to the shared wiki — a living knowledge graph of typed facts, per-entity append-only logs, and LLM-synthesized briefs committed under an archivist identity.

## Analogy

Think of it as a shared whiteboard for AI teammates, where everyone writes but nothing gets erased between shifts.

## Why it's emerging now

Multi-agent systems are mainstream in 2026, but team memory remains unsolved: each agent loses shared context on every session reset. Wiki Layer names the missing piece — a git-backed markdown store all agents can read and write, promoted from private notebooks. WUPHF shipped a working implementation and coined the term publicly on April 25.

## Related terms

- *parent:* llm-wiki
- *related:* agent-harness
- *parent:* context-engineering
- *related:* agents-md
- *related:* managed-agents
- *related:* workspace-agents
- *parent:* agent-team
- *related:* context-rot
- *competitor:* vault-context
- *competitor:* RAG
- *related:* personal-ai
- *related:* ai-workspace

## Sources

1. [WUPHF — Show HN: A Karpathy-style LLM wiki your agents maintain](https://news.ycombinator.com/item?id=47899844)
2. [nex-crm/wuphf — GitHub (reference implementation)](https://github.com/nex-crm/wuphf)
3. [Karpathy — llm-wiki.md gist (origin pattern)](https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f)
4. [MindStudio — LLM Wiki vs RAG](https://www.mindstudio.ai/blog/llm-wiki-vs-rag-knowledge-base)
5. [Intelligent Living — Karpathy LLM Wiki markdown knowledge base](https://www.intelligentliving.co/karpathy-llm-wiki-markdown-knowledge-base/)
6. [HN flagship thread — LLM Wiki idea file (296 pts)](https://news.ycombinator.com/item?id=47640875)

---
_Generated by EarlyTerms · https://earlyterms.com/term/wiki-layer_