Nicheloom

Market intelligence for builders — see what's gaining traction before it's crowded.

AI memory with biological decay (52% recall)

Details

External ID
47914367
Source
HN
Company
—
Product
AI memory with biological decay (52% recall)
Website domain
github.com
Launched
April 26, 2026
Cohort
—
Upvotes
98
Upvotes percentile
0.8920308483290489
Tags
—
Fetched at
Sept. 7, 2026, 9:26 p.m.
Updated at
Sept. 7, 2026, 9:26 p.m.

Description

Most RAG setups fail because they treat memory like a static filing cabinet. When every transient bug fix or abandoned rule is stored forever, the context window eventually chokes on noise, spiking token costs and degrading the agent's reasoning.This implementation experiments with a biological approach by using the Ebbinghaus forgetting curve to manage context as a living substrate. Memories are assigned a "strength" score where each recall reinforces the data and flattens its decay curve (spaced repetition), while unused data eventually hits a threshold and is pruned.To solve the "logical neighbor" problem where semantic search misses relevant but non-similar nodes, a graph layer is layered over the vector store. Benchmarked against the LoCoMo dataset, this reached 52% Recall@5, nearly double the accuracy of stateless vector stores, while cutting token waste by roughly 84%.Built as a local first MCP server using DuckDB, the hypothesis is that for agents handling long-running projects, "what to forget" is just as critical as "what to remember." I'd be interested to hear if others are exploring non-linear decay or similar biological constraints for context management.GitHub: https://github.com/sachitrafa/cognitive-ai-memory

Enrichment

Theme
specialized AI models and agent reasoning tools
Vertical
Horizontal
Function
Agent / copilot
Audience
Developer
AI stance
AI-native
Project type
Hobby / open-source project
Normalized one-liner
ai memory system with biological decay
Manually corrected
False

Could you build this?

Yes This is an experimental RAG retrieval strategy that applies mathematical decay scoring (e.g., Ebbinghaus forgetting curve equations) to vector/metadata database queries. Implementing custom decay scoring functions over an existing vector database (like Chroma or pgvector) is easily built with an AI assistant.

Discussion

20 comments analyzed.

Competitors mentioned: Google Research's Titans+Atlas, Claude Code's auto-memory, LongMemEval dataset benchmarks, RAG systems

Concerns raised: Wall clock decay penalizes users taking vacations, Memory systems in general may not improve outcomes for software work, Risk of comingling unrelated projects without proper scoping, Markdown files bloat and require manual maintenance, Every product launching long-term memory systems lacks clean breakthrough

Feature requests: Spaced repetition algorithm (Anki-like) for keeping relevant memories fresher, Per-fact files with tiny always-loaded index for on-demand loading, Filter on what gets written to distinguish globally vs locally-relevant information, Multi-step task context injection without per-turn recall triggering

Competitors

Other products that read as similar to this one — 155 launches clear the similarity bar, closest 8 shown.

Attention rank: #17 of 156 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).

Launched 171 days after the earliest competitor.

Other launches for this product

Same idea, different domain

Nobody's really built a agent / copilot tool for Agriculture yet.