Most AI workflows today still revolve around one massive prompt.
Bigger context. Bigger instructions. Bigger prompt engineering tricks.
Macey Baker (
@macebake) and Baruch Sadogursky (
@jbaruch) are joining AI Native DevCon London 2026 to argue for a completely different approach: stop treating agents like chatbots and start treating them like software systems.
Their workshop, Donβt Write Prompts, Write Software, explores how modular skills can replace fragile prompt-heavy workflows with reusable, testable building blocks for agent behavior.
Because once teams start scaling AI workflows across real projects, giant prompts stop being maintainable very quickly.
What this workshop covers
β’ How to break agent workflows into composable behavioural units
β’ Why skills are more reliable than giant all-in-one prompts
β’ Techniques for testing, refining, and evaluating agent behaviour
β’ How to package reusable skills into plugins and workflows
β’ Practical approaches to making AI systems easier to debug and improve over time
Macey brings experience spanning startups, ML systems, and AI-native developer workflows, while Baruch has spent decades helping shape developer tooling, DevRel, and platform engineering communities long before any of this was fashionable.
Together, theyβll walk through how teams can move from ad-hoc prompting to structured systems engineering for agents.
If youβve started realizing that prompt engineering alone doesnβt scale, this workshop is probably for you.
Join us in London or online:
tessl.io/devcon/ (use AIND-X-BB-20 for 20% discount)