Making agentic workflows trustworthy and verifiable with a custom DSL

ℹ️ EPISODE INFO

Released

May 22, 2026

Duration

29m

Channel

Claude

Watch on YouTube

▶ Open ↗

📝 EPISODE DESCRIPTION

System design of agentic research assistant built unconventionally: one component outputs plan in custom Turing-incomplete programming language, another interprets it, quiver of models executes concrete tasks. Architectural choices as concrete instantiations of company values.

🧾 EPISODE SUMMARY

In this episode of Claude, Making agentic workflows trustworthy and verifiable with a custom DSL explores elicit’s DSL makes agent workflows legible, faithful, and cacheable execution The talk argues that identical outputs should not be equally trusted because the internal mechanism (model choice, tooling, critique, token spend) materially changes reliability and risk.

🔗 RELATED EPISODES

How we Claude Code

Agent Battle: Mine the most diamonds in 45 minutes

Evals for taste: Hill-climbing a slide-generation agent

Agents that remember

Tool, skill, or subagent? Decomposing an agent that outgrew its prompt

How AirOps chases friction to build AI products with Claude