No RAG or Context Window: New architecture treats LLM as just a language center
LowDistribution3995 · reddit · 2026-08-29
The author proposes a new backend agent memory architecture aiming to eliminate context windows and RAG. It uses dual perpendicular graphs and dual cyphers for real-time auto-cross-referencing, outputting embedded state parameters instead of natural language memory chunks. Tests with a GGUF adapter and a 0.6B Qwen3 model show accurate sentence output without natural language prompts. Theoretically, it maintains conversations without external caches, treating the LLM solely as a language center while routing actionable outputs at the system level.
More from coding & agent
- Forcing bot names isn't dumb — delegate them: a Grok Bot orchestration workflow — omarsar0 · 2026-08-30
- Codex nearly one-shots a 20-year-old PHP codebase rebuild in modern JavaScript — cnakazawa · 2026-08-30
- Ex-Engineering Director: Coding with LLMs is building the system around the system — viksit · 2026-08-30
- Are there real business use cases for true agentic systems beyond customer support? — AdGrouchy7150 · 2026-08-30
- Bot Mesh adds transparency mode: humans can see through public bots — Daniel_Farinax · 2026-08-30
- LongHorizon-Harness: external task state + audit loop triples OSWorld agent scores — rohanpaul_ai · 2026-08-30