AI/ML●●Solid
I benchmarked how good LLMs are at proofreading English
Agent loop proofreading evals where HELM and LMSys are too generic.
Solve My ProblemShip It
artursapek
321mo ago
Back to browse
GitHub Repository
One brain. Many bodies. Orchestration framework for embodied AI built on SCP. LangGraph for physical systems. Zero HTTP between bodies.
4 starsJavaScript
SCP – A protocol that drops LLM API calls to zero in 60fps physics loop
by srk0102·Apr 14, 2026·1 point·0 comments
AI Analysis
●●●BangerBig BrainWizardry
Subsumption Architecture revival cuts LLM calls with pattern cache misses.
Strengths
- •Pattern cache means LLM only invoked on unseen states, not every frame
- •Token budget trimming by event priority keeps prompts within limits
- •Zero external dependencies beyond scp-protocol, uses node:* built-ins
Weaknesses
- •Niche audience - only matters if you're building embodied AI systems
- •No safety layer or planner, just validation and dispatch
Category
Target Audience
Robotics developers, embodied AI researchers
Similar To
LangGraph · CrewAI · AutoGen
Post Description
I was building with MuJoCo and realized passing a 60fps state to an LLM is impossible. So I implemented Rodney Brooks' 1986 Subsumption Architecture for modern LLMs.
Similar Projects
Data●Mid
Designing an API for LLMs, Not Humans
Interesting methodology but this is a blog post, not a product you can use.
Big Brain
RileyHilliard
302mo ago
Security●●●Banger
Veil a Drop-in PII redaction proxy for any LLM API
Stops zero-width Unicode bypasses that break standard PII filters before LLM calls.
WizardrySolve My Problem
A5omic
202mo ago
AI/ML●●Solid
Qwen Meetup, Function Calling Harness, turning 6.75% to 100%
Compiler-level validation turns Qwen's 6.75% structured output success rate into 100%.
Big BrainNiche Gem
samchon
102mo ago
AI/ML●●Solid
MUP – Stop making LLMs rebuild UI from scratch every time
LLMs operate pre-built UI components instead of burning tokens regenerating Chart.js every time.
Big BrainZero to One
Ricky_Tsou
102mo ago
AI/ML●●Solid
An agentic loop for time-series forecasting using sktime and MCP
MCP proxy prevents context-window bloat better than naive LLM forecasting.
Big BrainDark Horse
silkpeak
102mo ago