Architectural breakdowns, AI agent systems, Model Context Protocol (MCP) servers, web engineering, and authentic founder notes published from Junagadh, Gujarat, India.
Journal & Field Notes
Autonomous Code Review & QA Swarms: How Multi-Agent AI Pipelines Eliminate Bugs Before Production
How to build an automated multi-agent CI/CD pipeline: orchestrating AST linters, security auditors, test generators, and auto-patching agents to protect...
Read the story →Best AI Agent Frameworks August 2026: Ranked by Alice Labs
Alice Labs ranked 10 AI frameworks in August 2026 — LangGraph #1, MAF #2, ADK #5. From Junagadh, the production lens that beats star counts.
Read the story →Claude Agent SDK: 5-Level Hierarchical Subagents Ship
Claude Agent SDK ships 5-level hierarchical subagents with deepest MCP — from Junagadh, when to use it over LangGraph and hooks gain. Includes measured
Read the story →DeepSeek Harness: 95K Stars in 48h — Everything Is a Plugin
DeepSeek Harness hit 95K stars in 48h with everything-is-a-plugin runtime. From Junagadh I tested its hot-swap, audit log and plugin gaps before you ship.
Read the story →Deploy Google ADK to Cloud Run & Vertex AI: Ship Sunday
Deploy Google ADK from adk web localhost to Cloud Run & Vertex AI in one Sunday night — eval harness, cost ~$0.002 per call and the 7-step checklist that ships.
Read the story →Frontier Model Routing 2026: Mistral vs Gemini vs Claude
Routing Mistral vs Gemini vs Claude in 2026 with thinking budgets 0–64K — from Junagadh, the SLM router that cut a legal-tech bill 85% without loss.
Read the story →Google ADK Multi-Agent Patterns: 8 Designs That Ship
8 Google ADK multi-agent patterns that ship in 2026 — Sequential, Parallel, Loop, Hierarchical — with code, latency wins and the 4pm state bug fix.
Read the story →Google ADK vs LangGraph vs CrewAI: Which to Pick in 2026
ADK vs LangGraph vs CrewAI in 2026? I built the same pipeline in all three from Junagadh — speeds, costs, tracing and the pick rule that saves rewrites.
Read the story →Gujarat SME Automation: WhatsApp + UPI + n8n in 30 Days
Gujarat SME automation with WhatsApp Business, UPI and n8n pays in 30 days — from Junagadh, the stack that cuts COD and recovers COD at scale.
Read the story →Hermes Agent 234K Stars: Self-Improving AI That Learns
Hermes Agent hit 234K stars with a self-improving loop that creates skills from experience. From Junagadh, $5 VPS vs local test and what persists.
Read the story →LlamaIndex Workflows 1.0: Event-Driven Engine Left RAG
LlamaIndex Workflows 1.0 is the standalone event-driven engine — @step, typed events, no RAG dep — from Junagadh the smallest orchestrator that scales.
Read the story →Mastra 300K Weekly: TypeScript Agent Framework Scales
Mastra hits 300K weekly npm as TypeScript-first agent framework — from Junagadh, when TS beats Python and the Studio that ships it. The ledger that travels f...
Read the story →Microsoft Agent Framework 1.0: Harness Goes GA
Microsoft Agent Framework 1.0 went GA April 2026 — harness, hosted agents and 40-step brake. From Junagadh, what to migrate and what to lab.
Read the story →OpenAI Agents SDK GA: Sandbox Harness for 7 Providers
OpenAI Agents SDK went GA March 2026 with sandbox harness across 7 providers — from Junagadh, lightweight delegation that still governs. The ledger that trav...
Read the story →Pydantic AI 2.0 + Harness: Type-Safe Agents That Ship
Pydantic AI 2.0 harness-first with type-safe agents — from Junagadh, the schema-contract pattern that holds hallucination under 0.3% in production.
Read the story →