NERDs
Entity-centered long-term memory for LLM agents
Details
- External ID
- 47277446
- Source
- HN
- Company
- —
- Product
- NERDs
- Website domain
- nerdviewer.com
- Launched
- March 6, 2026
- Cohort
- —
- Upvotes
- 13
- Upvotes percentile
- 0.6660516605166051
- Tags
- —
- Fetched at
- Sept. 7, 2026, 9:26 p.m.
- Updated at
- Sept. 7, 2026, 9:26 p.m.
Description
Long-running agents struggle to attend to relevant information as context grows, and eventually hit the wall when the context window fills up.NERDs (Networked Entity Representation Documents) are Wikipedia-style entity pages that LLM agents build for themselves by reading a large corpus chunk-by-chunk. Instead of reprocessing the full text at query time, a downstream agent searches and reasons over these entity documents.The idea comes from a pattern that keeps showing up: brains, human cognition, knowledge bases, and transformer internals all organize complex information around entities and their relationships. NERDs apply that principle as a preprocessing step for long-context understanding.We tested on NovelQA (86 novels, avg 200K+ tokens). On entity-tracking questions (characters, relationships, plot, settings) NERDs match full-context performance while using ~90% fewer tokens per question, and token usage stays flat regardless of document length. To highlight the methods limitation, we also tested it on counting tasks and locating specific passages (which aren't entity-centered) where it did not preform as well.nerdviewer.com lets you browse all the entity docs we generated across the 86 novels. Click through them like a fan-wiki. It's a good way to build intuition for what the agent produces.Paper: https://www.techrxiv.org/users/1021468/articles/1381483-thin...
Enrichment
- Theme
- Vertical
- Horizontal
- Function
- Model & infra
- Audience
- Developer
- AI stance
- AI-native
- Project type
- Commercial product
- Normalized one-liner
- entity-centered memory for llm agents
- Manually corrected
- False
Could you build this?
Yes It is an agentic document-processing and memory framework that extracts entity knowledge graphs and generates markdown/wiki pages chunk-by-chunk using standard prompt engineering and vector/document databases.
Discussion
5 comments analyzed.
Concerns raised: Auto-regressive error propagation when LLM starts with incorrect information, Early incorrect assumptions about relationships propagating through entity graph, System stuck with incorrect entity titles when introduced early, Inconsistent entity linking when same entity appears under different names, Potential catastrophic forgetting if agent revises NERDs
Feature requests: Tool to rename/update entity titles after creation, Agent capability to review entities for contradictions and search source material, Belief revision support for correcting early incorrect assumptions, Test performance on papers, conversations, and other text formats
Competitors
Other products that read as similar to this one — 229 launches clear the similarity bar, closest 8 shown.
Attention rank: #72 of 230 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).
Launched 127 days after the earliest competitor.
- A file-based agent memory framework that works like skill · hn · 2026-01-06 · 11 upvotes · similarity 0.47
- genpark-graph-rag-entity-relationship-extractor-skill · github · 2026-09-28 · 7 upvotes · similarity 0.46
- ChatIndex · hn · 2025-11-26 · 17 upvotes · similarity 0.45
- I built "AI Wattpad" to eval LLMs on fiction · hn · 2026-02-03 · 32 upvotes · similarity 0.44
- Unify memory across agents and improve context rot, written in Rust · hn · 2026-04-05 · 5 upvotes · similarity 0.44
- AI-Augmented Memory for Groups · hn · 2025-12-16 · 10 upvotes · similarity 0.44
- Honcho · hn · 2026-01-27 · 8 upvotes · similarity 0.43
- Collabmem · hn · 2026-04-11 · 11 upvotes · similarity 0.43
Other launches for this product
- No other launches for this product.
Same idea, different domain
Nobody's really built a model & infra tool for Fintech yet.