Education●●●Banger
DAG-based Kanji learning through components
Visualizes kanji structure as a recursive DAG when WaniKani only uses static mnemonics.
Big BrainNiche GemSlick
barisozmen
501mo ago
Back to browse
GitHub Repository
Proof of concept for using LLMs to generate an application by following a recursive graph decomposition of the application's architecture.
1 starsPython
Plan-Graph based code generation with LLMs
by ag_rin·May 16, 2026·2 points·0 comments
AI Analysis
●MidShip ItBold Bet
Graph decomposition for LLM code gen, but takes an hour for a calculator app.
Strengths
- •Recursive graph decomposition avoids whole-project context overload for LLMs.
- •Structured guardrails with deterministic tooling is the right architectural instinct.
Weaknesses
- •Token-inefficient: over an hour runtime for a 5-component calculator app.
- •Raw scaffolding, 0 stars, author admits it's just a research experiment.
Category
Target Audience
Developers experimenting with LLM code generation workflows
Similar To
Cursor · Continue · Aider
Post Description
Creates a implementation plan using a graph of components instead of markdown docs and then implements each component in a loop before connecting them. My target was to figure out how to make LLM code more maintainable and a computation graph seemed to be the way. What do you think, any merit to the idea and has anyone tried a similar path?
Similar Projects
Developer Tools●●Solid
Rlm-codelens – Codebase intelligence with recursive language model
Graph-aware RLM decomposition beats context-window limits; but Codeium/Sourcegraph Cody solve this already.
Big BrainWizardry
wizmlveteran
213mo ago
AI/ML●●Solid
Replaced Neo4j with pure vector search for Graph RAG
Graph RAG without Neo4j — pure vector search beats HippoRAG on multi-hop benchmarks.
Big BrainDark Horse
zhangchen
202mo ago
Developer Tools●Mid
Math-spec-driven LLM skill for complex system specs
Math-spec approach for LLM-generated code, but lacks working examples and doesn't solve the reasoning-accuracy problem.
Big Brain
ben8t
103mo ago
AI/ML●Mid
Sovereign Suite – A Recursive Logic Framework for AI Governance
The repo frames a neat, mathematically flavoured approach: three complementary auditors (generation, explanation, adversarial skepticism) scored with MSE and even bounded using Cauchy's integral formula — a bold, uncommon toolkit for model auditing. It comes with a white paper and a self-audit certificate, but the landing page and repo contents don't make the practical integrations or benchmarks obvious, so it's interesting research that still needs reproducible pipelines and clear evaluation to win broader adoption.
Big BrainNiche Gem
badatchess
203mo ago
SaaS●●Solid
Motionode – Cursor for Technical Planning
They split the work smartly: an LLM extracts deliverables from messy docs, then a deterministic 'physics' engine computes dependencies, capacity-based timelines, and a defendable critical path. Client-ready SOW PDFs, a scope-change ledger, and on-the-fly what-if scenarios feel immediately useful; I'd want to see proof it integrates cleanly with existing PM tooling and handles truly noisy inputs.
Big BrainSolve My ProblemSlick
oscarcaldera
203mo ago