Blog
Technical knowledge made accessible. Articles on AI, automation, development and the Cadences ecosystem.
The Claude Code System Prompt: 30,000 Tokens of Instructions That Define an Engineering Agent
Complete analysis of the Claude Code system prompt: 30K+ tokens orchestrating tools, security and reasoning. Lessons applicable to any AI agent.
Synapse Agents vs Claude Code Coordinator: Two Agent Orchestration Philosophies
Technical comparison between multi-agent orchestration in Synapse and the coordinator model in Claude Code. Sub-agents, task delegation and communication patterns.
RAG with LLM Enrichment: From Plain Text to Semantic Intelligence
How we improved our RAG chatbot accuracy by ~40% using HyDE (Hypothetical Document Embeddings), DeepSeek V3 enrichment and semantic metadata. Real implementation for under $0.05.
RADIA: AI Diagnostics in Fundoscopy & Ultrasound — From Retinoblastoma to Soft Tissue Sarcoma
How RADIA analyzes fundoscopy images and Doppler ultrasound with multimodal AI: multicentric retinoblastoma detection, papilledema, ischemic retinopathy, and hypervascularized soft tissue masses.
NutriNen Context Engine vs Claude Code: Two AI Context Management Strategies
How NutriNen and Claude Code solve the same context problem with opposite approaches: specialized RAG vs massive context window. Real performance metrics.
Cadences Permissions vs Claude Code ACL: Access Control Models in AI Platforms
Comparative analysis of permission models: Cadences multi-tenant RBAC vs Claude Code tool ACL. Two security philosophies for AI platforms.
Magic Docs vs Synapse Auto-Documentation: AI That Documents Your Code Automatically
Comparison between Claude Code's Magic Docs and Synapse auto-documentation. Two approaches to solving technical documentation with generative AI.
The Tool Permission System in Claude Code: Security, Control and Safe Execution
How Claude Code manages tool permissions: allowedTools, progressive trust model and safe execution. Patterns applicable to any autonomous agent.
The Dual Life of Claude Code: Anthropic's Internal Build vs External Extension
Claude Code has two modes of operation: one internal optimized for Anthropic and another for external developers. Analysis of the differences and what we can learn.
Context Compression in Claude Code: How to Maintain 200K Tokens Without Losing Critical Information
How Claude Code compresses long conversations without losing critical information. Auto-compaction, hierarchical summarization and intelligent context window management.
Memory Systems in Claude Code: CLAUDE.md, Project Files and Persistent Memory
Complete analysis of the three memory levels in Claude Code: CLAUDE.md, project files and session context. How to implement persistent memory in AI agents.
Hidden Features of Claude Code: /dev/stdin Notation, Ultra-Thinking and Undocumented Tricks
The hidden features of Claude Code that aren't in the official documentation: /dev/stdin, ultra-thinking mode, tool result injection and more advanced tricks.
Claude Code vs Synapse Studio: Two Multi-Agent Orchestration Philosophies
A real technical comparison between Anthropic's CLI and the Cadences engine. Execution pipelines, permissions, hooks, context management and multi-agent coordination. What each system does best and why.
Parallel Execution on the Edge: Self-Fetch Chain and Unlimited Orchestration
How to run long-running multi-agent tasks on Cloudflare Workers without hitting time limits. Self-Fetch Chain pattern, timeout cascading and real parallel execution.
Edge-Native vs Traditional Cloud: A Practical Guide to Choosing Your AI Stack
Azure, AWS, GCP or Cloudflare Workers — the architecture patterns are the same. Service mapping, cost comparison, semantic search and when to choose each platform.
Semantic Search & RAG: How Your AI Understands Before Answering
What embeddings are, how semantic search works, and why RAG is the pattern that eliminates AI hallucinations. Real implementation with Vectorize, Workers AI and DeepSeek.
From Idea to Product: NutriNen Anatomy — AI, SEO & Native App with No Team
How to identify a market gap, validate it with 29 free SEO tools, build a chatbot with 33 function calling tools and package everything as a native Android app.
The Agentic SDLC Without Hype: Structure, Not Agent Count
Beyond the buzzword: what an AI agent-driven development lifecycle really needs to work in production. Real spec-driven approach, disciplined orchestration and Synapse Studio as a case study.
Security, GDPR & Compliance: How We Protect Every Byte
Turnstile, CSP, OAuth PKCE, per-tenant isolation, cookie consent, right to erasure and premium feature guards — Cadences multi-layer security explained.
Integrations and APIs: Connect Cadences with Everything
Google, Twilio, ElevenLabs, Stripe, WhatsApp, Cloudflare — how Cadences connects with 15+ external services through REST APIs, webhooks and OAuth flows.
ML Without Cloud: From Browser to Desktop with TensorFlow.js, ONNX and a Local OpenAI API
A complete ML pipeline without a server: TensorFlow.js in-browser, Transformers.js with ONNX in Electron, pure JS gradient descent, OpenAI-compatible localhost API, browser↔desktop sync and federated learning.
Desktop Apps with Electron: CadencesLab, Audio Hub & WhatsApp Agent
How we built 3 desktop apps with Electron: offline-first with IndexedDB, voice synthesis with ElevenLabs, WhatsApp automation with Playwright and cross-platform packaging.
From Sensor to Headset: IoT Hub & Audio Hub as Desktop Agents Orchestrated by Cadences
Two Electron apps that register as agents in Cadences — IoT Hub (40+ sensors, MQTT, Serial, cameras with ffmpeg) and Audio Hub (Whisper STT, 5-engine TTS cascade, Bluetooth).
Synapse: Your Business Data Brain
Inside Synapse Studio — the central nervous system that organizes AI agents, tasks, events and context with gamification, leaderboards and autonomous execution.
React in cadences.app: Architecture of 80+ Components Without Abstraction Libraries
How we built a SPA with pure React 19: custom SVG workflow editor, schema-driven form system, data tables, 23 custom hooks and a 4,750-line persistence hook.
Conversational Voice AI: From Robotic IVRs to Natural Conversations
How Cadences connects Twilio, ElevenLabs and Durable Objects to create voice agents that hold natural phone conversations in 28+ languages.
AI Agents in Cadences: From Prompt to Autonomous System
How a simple text becomes an agent with personality, memory, tools, and the ability to make decisions on its own. Agent library, scheduling and multi-provider execution.
Multi-Tenant Architecture: How Cadences Scales for Thousands of Organizations
A technical deep dive into Cadences' edge-native architecture. D1 per organization, Durable Objects, total data isolation and cost model.
No-Code Automation: Workflows that Work for You
Design complex automation flows without writing a single line of code. Triggers, conditions, actions, webhooks and how it compares to Zapier and Make.
Intelligent CRM: AI-Powered Client Management
How Cadences transforms contact management with automatic enrichment, smart search, predictive deal scoring and AI-powered activity timelines.
Storefronts Explained: Professional Websites with Superpowers
Understand what Cadences Storefronts are and why they're not "just a static website". CMS, CRM, integrated AI and on-demand recompilation.
Generative AI at Work: From Theory to Real Practice
How to implement generative AI in your company without dying trying. Real use cases, automated workflows and the future of human-AI hybrid work.
Introduction to Cadences: The Platform that Unifies Your Business
Discover how Cadences integrates CRM, automation, AI and professional websites into a single platform. A complete guide to understanding the ecosystem.