Continual Learning with .md
Details
- External ID
- 47757552
- Source
- HN
- Company
- —
- Product
- Continual Learning with .md
- Website domain
- github.com
- Launched
- April 13, 2026
- Cohort
- —
- Upvotes
- 34
- Upvotes percentile
- 0.8046272493573264
- Tags
- —
- Fetched at
- Sept. 7, 2026, 9:26 p.m.
- Updated at
- Sept. 7, 2026, 9:26 p.m.
Description
I have a proposal that addresses long-term memory problems for LLMs when new data arrives continuously (cheaply!). The program involves no code, but two Markdown files.For retrieval, there is a semantic filesystem that makes it easy for LLMs to search using shell commands.It is currently a scrappy v1, but it works better than anything I have tried.Curious for any feedback!
Enrichment
- Theme
- ML inference and model optimization
- Vertical
- —
- Function
- —
- Audience
- —
- AI stance
- —
- Project type
- —
- Normalized one-liner
- —
- Manually corrected
- False
Could you build this?
Yes The project is literally described as involving 'no code, but two Markdown files' along with basic shell commands for semantic file search. Anyone can create or adapt this setup in an afternoon with an AI assistant.
Discussion
20 comments analyzed.
Competitors mentioned: LLM Wiki (Karpathy's approach), Embedding-based knowledge bases, MCPTube
Concerns raised: Embedding-based methods struggle matching user queries to stored information, Risk of memory pollution and false information persisting (Memento Effect), No continuous running or LLM ablation studies yet (MVP stage), Lack of benchmark measuring how well memories contribute to future interactions, Query reformulation with embedding-based methods is tricky with chat history conditioning
Feature requests: Implement removal/pruning mechanism to refactor filesystem daily, Create benchmark to measure memory contribution to future interactions, Add continuous running capability and LLM comparison studies, Support for native video file reading on Gemini without embeddings
Competitors
Other products that read as similar to this one — 130 launches clear the similarity bar, closest 8 shown.
Attention rank: #32 of 131 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).
Launched 165 days after the earliest competitor.
- A file-based agent memory framework that works like skill · hn · 2026-01-06 · 11 upvotes · similarity 0.52
- Llm.sql · hn · 2026-04-24 · 8 upvotes · similarity 0.44
- Breathe-Memory · hn · 2026-03-26 · 6 upvotes · similarity 0.43
- note.md · ph · 2026-06-26 · 279 upvotes · similarity 0.43
- Learning Rust by writing a Markdown to HTML compiler · hn · 2026-07-29 · 25 upvotes · similarity 0.42
- Fixing LLM memory degradation in long coding sessions · hn · 2025-11-27 · 5 upvotes · similarity 0.42
- Persistent Memory for Claude Code (MCP) · hn · 2026-01-10 · 7 upvotes · similarity 0.42
- A local-first memory store for LLM agents (SQLite) · hn · 2025-12-14 · 48 upvotes · similarity 0.42
Other launches for this product
- No other launches for this product.