Spec Driven Development with Spec Kitty
Spec Kitty turns intent into reviewable specs, governs execution across the tools developers already use, and leaves an auditable delivery record from first decision to merge.
Agents are invading modern software development. The gap between what we whisper into the black box (our Intent) and what we actually ship (the 100K LOC we didn't read) can be costly. Software teams are asking the question "Does this work on my legacy codebase with 3M lines in an ancient language?" This talk introduces spec-driven development: a methodology where clear, executable specifications become the north star of your development process, and the cognitive load of reviewing agentic output shifts left.
How does Spec Kitty streamline this approach, and what limitations have we found regarding old code, complex code, or working in restricted and regulated environments? How can software teams safely adopt a new agentic-guided paradigm safely, and ship with confidence? Whether you're working on a team of one or managing distributed developers, you'll learn how to:
* Write specifications that guide implementation and prevent ambiguity
* Create governance Doctrine that keeps agents on the rails
* Direct human intelligence into the part of the software development lifecycle where it matters the most
* Safely start an agentic loop that can run for hours or days and produce a result you're comfortable with
I'm the creator of Spec Kitty, an MIT licensed Python CLI, and I'll show you live examples of how it is being used today, from solo developers to Fortune 500 companies launching critical initiatives.
Licensed to the public under http://creativecommons.org/licenses/by/4.0