SpecDD, an open-source framework for specification-driven development, is now available for teams that want AI coding to become more than an impressive demo. Built for organizations already using AI to create software, SpecDD gives product, QA, support, operations, and engineering teams a shared way to preserve intent before AI turns incomplete requirements into polished but wrong software. SpecDD open-source framework is available at https://specdd.ai.
AI can now write code faster than many organizations can explain what the code is supposed to mean. That is the new bottleneck.
In most teams, intent is scattered across planning, customer knowledge, operational judgment, review comments, and memory. By the time it reaches an AI coding tool, it is often reduced to a narrow prompt. The tool then does exactly what it was built to do: it generates something plausible.
Plausible is no longer good enough. The next step is specification-driven development for AI: turning intent, requirements, boundaries, and completion criteria into reusable context before code is generated.
SpecDD, a specification-driven development framework for AI-assisted software teams, was created for the next phase of software delivery, where the goal is not just faster code, but faster correct delivery. It helps teams capture what must happen, what must not happen, and what “done” means in a durable form that humans can review and AI tools can use.
“The problem is not that AI cannot write code,” said Matīss Treinis, creator of SpecDD. “The problem is that we keep asking AI to build from fragments of intent. SpecDD gives the intent a place to live before the work begins.”
In private internal testing, the impact was immediate:
- SpecDD reduced observed correction loops from roughly 10-20 prompt-and-correction cycles to 1-2 cycles for comparable feature and service-class work on commercial code.
- Some comparable work moved from multi-day iteration to same-afternoon completion.
- In out-of-domain experiments, where software was built without prior team expertise in the target domain, SpecDD reduced observed time to live by 75%-89% compared with unstructured AI-assisted coding workflows.
- Compared with root-level documentation workflows, SpecDD reduced observed time to deliver by 55%-67% in the same out-of-domain test pattern.
The lesson is clear: structured intent changed the shape of the work. AI needed less steering, review had a clearer target, and usable delivery arrived faster.
“This was the moment SpecDD became obvious to me,” Treinis said. “The same class of work stopped bouncing through double-digit correction loops. Out-of-domain work that normally drifts through vague prompting became much more direct. That is the difference between AI feeling impressive in a demo and AI becoming useful in real delivery.”
SpecDD is built around a simple shift: stop treating the prompt as the source of truth. The prompt is too temporary, too narrow, and too easy to lose. In AI-assisted delivery, the source of truth needs to survive across sessions, tools, contributors, reviews, and product changes.
That shift matters for every team adopting AI coding:
- Founders can move faster without letting the product drift away from the original idea.
- Product teams can preserve business rules before they become ambiguous implementation choices.
- QA teams can make known edge cases part of the delivery contract instead of rediscovering them at the end.
- Support and operations teams can capture real workflow constraints that rarely fit cleanly into tickets.
- Engineering leaders can scale AI coding without turning senior review into the place where missing context is reconstructed again and again.
SpecDD is especially useful anywhere plausible mistakes are expensive: onboarding, billing, permissions, account recovery, internal tools, customer-facing error states, regulated workflows, support-sensitive behavior, and systems where a small misunderstanding can create a large correction cycle.
“There is a lot of talk about AI replacing software work,” Treinis said. “The winning strategy is not replacing the people with the best judgment. It is making them dramatically more productive. SpecDD turns their intent into a reusable context for AI, helping companies increase development velocity while maintaining, and in many cases improving, delivery quality.”
SpecDD does not remove product judgment, QA, engineering review, or human accountability. It makes those functions more useful by giving everyone the same contract before AI-generated work runs ahead. The goal is not slower AI. The goal is AI that can move quickly without leaving the organization to clean up misunderstood intent afterward.
Teams can try SpecDD immediately by choosing one feature, workflow, or high-risk area where intent is currently scattered across delivery tools and team memory. Capture the intent first, then let AI build against it.
SpecDD is available at https://specdd.ai and is released under open-source Apache License 2.0.
About SpecDD
SpecDD is an open-source framework for specification-driven development in AI-assisted software projects. It helps teams preserve product intent, delivery constraints, edge cases, and completion criteria in a shared form that business stakeholders, engineering teams, and AI tools can use. SpecDD is designed to reduce rework, improve alignment, and help organizations scale AI-assisted development without losing sight of what the software is supposed to do.