Notes on architecture, AI, and the trade-offs nobody writes about.
Long-form pieces on multi-agent systems, MLOps, enterprise integration, and the engineering trade-offs that decide whether software actually survives production.
The End of the Shopping Cart: Agentic Commerce (A2A) and the New Operating System for Startups
We are transitioning from 'Search and Click' to 'Purpose and Execute.' As commerce accelerates to machine speed, the hidden crisis is that companies still operate at spreadsheet speed. To build for the agentic economy, the startup itself must become agentic.
AI Can Generate Code — But Only Observability Can Tell It If It Works
As AI generates more of our code, observability stops being a debugging tool for humans and becomes the communication layer between production systems and intelligent agents. Without runtime visibility, AI-generated systems remain static artifacts — they cannot know if their output actually works.
KafkaJS Works… Until It Doesn't: Lessons from Real-World Streaming Architectures
When KafkaJS meets enterprise scale: lifecycle hell, backpressure starvation, observability gaps, configuration spaghetti, and error recovery. Five hidden cracks that drive most production incidents — and the case for a Kafka runtime that behaves like a first-class citizen of the NestJS ecosystem.
Is Your NestJS Code a Tangled Mess? It's Time to Tame the Spaghetti Monster
Once-clean NestJS services devolve into nested if/else, manual state, and tightly-coupled dependencies. The fix is not more code — it's a clean separation between the workflow definition (what should happen) and the listeners (how it happens).
Error Detection in Enterprise Software: Time, Costs, and Challenges
Developers lose half their time debugging. Downtime costs thousands per minute. MTTD and MTTR remain stubbornly high. This essay surveys the real cost of error detection in modern enterprise software, the typical log-analysis loop, and why this remains an open problem worth solving.
NestJS Workflow & State Machine Module
A deep technical paper on @jescrich/nestjs-workflow: design and architecture, declarative workflow definitions, action handlers, comparison with XState and Temporal, and real-world use cases — order processing, medical prescriptions on Kafka, and support tickets.
Transforming Spaghetti Code into Structured Workflows with NestJS Workflow & State Machine
Traditional workflow code scatters status checks across services until even minor changes feel surgical. A declarative state machine — with built-in fallback states, Kafka integration, and centralized logging — replaces that complexity with a single source of truth.
URN and the Adapter Pattern in Microservice Architectures
Microservices that share URNs instead of vendor-specific IDs gain a stable, location-independent contract. Combined with the Adapter Pattern, URNs let teams add vendors as plug-ins, encode tenant context for clean billing, and keep the canonical model free from external noise.
The Art of Software Architecture: Crafting for Excellence and Efficiency in Delivery
In a world where no delivery means no revenue, architecture stops being a backstage discipline and becomes a front-line business decision. Good architecture buys you agility, parallel delivery, and time-to-market — bad architecture quietly burns your runway.
Harnessing the Adapter Pattern in Microservice Architectures for Vendor Agnosticism
When every vendor speaks a different dialect, the Adapter Service Pattern moves the translation layer out of your business code and into a dedicated, swappable microservice. Combined with a canonical data model and a smart API gateway, it makes vendor churn an architectural non-event.