Your agent architecture has a half-life of 6 months

Highlights

00:30

Dan Farrelly believes the half-life of Agent architecture is only six months, with trend shifts forcing constant system rewrites.

03:06

He predicts that the execution layer is more important than architectural patterns, and a reliable event-driven engine can reduce migration costs.

04:13

He opposes reliance on fixed frameworks and advocates placing decision logic close to code for technology stack flexibility.

⭐⭐⭐⭐ 4.0

Dan Farrelly, CTO of Inngest, argues that agent architectures have a half-life of about six months because each new trend—RAG, ReAct, prompt chaining, orchestrator-workers, MCP, CLI—forces a complete rebuild.

Rather than chasing the next pattern, he advocates building with durable primitives: bringing decisions closer to code for stack flexibility and choosing an execution layer that enables fast iteration.

The talk focuses on identifying the core primitives common to every production agent and designing systems that outlast the churn.

Dan Farrelly, CTO of Inngest, argues that agent architectures have a half-life of about six months because each new trend—RAG, ReAct, prompt chaining, orchestrator-workers, MCP, CLI—forces a complete rebuild.

Rather than chasing the next pattern, he advocates building with durable primitives: bringing decisions closer to code for stack flexibility and choosing an execution layer that enables fast iteration.

The talk focuses on identifying the core primitives common to every production agent and designing systems that outlast the churn.

Your agent architecture has a half-life of 6 months — Dan Farrelly, CTO, Inngest

View Original