The open-source project 'spec' helps structure developer intent, shifting the focus from writing code to writing effective specifications for AI agents.
Developers often struggle with effective prompting and defining quality gates, leading to a divide between organizations with disciplined processes and those adopting AI naively.
Use tools to translate natural language brain dumps into structured Markdown and bulleted lists, which are the most effective formats for AI agents to follow.
Vague prompts like 'fix it' produce poor results regardless of the AI tool, and iterating on generated code is more efficient than constantly regenerating it from scratch.
Successful teams build inner-sourced libraries of effective prompts and specs to improve consistency, reduce token costs, and avoid overwhelming projects with massive pull requests.
A significant gap exists between the hype around agentic AI and the reality that most organizations are still at the very beginning of their adoption journey.
Don't wait for a perfect methodology; start experimenting now to find what works for you and view AI as an exoskeleton that enhances your skills.
36:59Patrick Schnell
58:38Chris Heilmann, Daniel Cranney & Julia Kordick
45:09Gunnar Grosch
43:05Kenneth Auchenberg & Christian Heilmann
59:16Ryan J Salva
57:43Lee Faus
31:39Rainer Stropek
44:20Cassidy Williams
Jobs that call for the skills explored in this talk.
Avanade
Berlin, Germany
IO Sphere
Charing Cross, United Kingdom
Mindrift
Be-IT
Manchester, United Kingdom
Avanade
München, Germany
83zero Ltd
Charing Cross, United Kingdom
