Graph-Oriented Generation
Beating RAG for Codebases by 89%
Details
- External ID
- 47280787
- Source
- HN
- Company
- —
- Product
- Graph-Oriented Generation
- Website domain
- github.com
- Launched
- March 6, 2026
- Cohort
- —
- Upvotes
- 12
- Upvotes percentile
- 0.6439114391143912
- Tags
- —
- Fetched at
- Sept. 7, 2026, 9:26 p.m.
- Updated at
- Sept. 7, 2026, 9:26 p.m.
Description
LLMs are better at being the "mouth" than the "brain" and I can prove it mathematically. I built a deterministic graph engine that offloads reasoning from the LLM. It reduces token usage by 89% and makes a tiny 0.8B model trace enterprise execution paths flawlessly. Here is the white paper and the reproducible benchmark.
Enrichment
- Theme
- ML inference and model optimization
- Vertical
- Horizontal
- Function
- Agent / copilot
- Audience
- Developer
- AI stance
- AI-native
- Project type
- Commercial product
- Normalized one-liner
- code generation for codebases
- Manually corrected
- False
Could you build this?
Partial The LLM integration is simple, but creating a deterministic static code analysis engine that accurately maps entire codebase control-flow graphs for RAG offloading requires compiler expertise.
What it would actually take: Requires parsing source code into control-flow graphs, call graphs, and data-dependency graphs using parsers like Tree-sitter or LSP language servers, then storing these in an efficient graph structure (e.g., NetworkX or Rust/petgraph). The difficult aspect is building robust static analysis algorithms to reliably trace execution paths and cross-boundary dependencies across dynamic languages without missing runtime connections. Requires expertise in static program analysis, graph algorithms, and formal verification.
Discussion
2 comments analyzed.
Competitors mentioned: Language-specific LSP MCP for agents, IDE code navigation tools
Concerns raised: No citations or references, Single testcase benchmark only, Lacks comparison to existing approaches, Questionable research paper quality, Agents using grep inefficiently vs IDE navigation
Competitors
Other products that read as similar to this one — 798 launches clear the similarity bar, closest 8 shown.
Attention rank: #261 of 799 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).
Launched 125 days after the earliest competitor.
- RAGless · hn · 2026-08-14 · 6 upvotes · similarity 0.51
- Graphify Labs: Knowledge Graph Engine for Enterprises 🧠 · yc · 2026-08-19 · 66 upvotes · similarity 0.51
- KillSwitch · hn · 2026-09-19 · 10 upvotes · similarity 0.50
- The world's least deterministic programming language · hn · 2025-12-12 · 5 upvotes · similarity 0.49
- rizzo-flow · github · 2026-09-21 · 455 upvotes · similarity 0.48
- 10x better performance from the Coding Harnesses with LLM-wiki · hn · 2026-06-18 · 16 upvotes · similarity 0.48
- Compresr – context compression for LLM pipelines and agents 🗜️ · yc · 2026-02-25 · 9 upvotes · similarity 0.48
- graph-engineering · github · 2026-09-09 · 21 upvotes · similarity 0.47
Other launches for this product
- No other launches for this product.
Same idea, different domain
Nobody's really built a agent / copilot tool for Agriculture yet.