Revolutionising AI Coding: Why GitHub’s Spec Kit Feels Like the Missing Piece

Not long ago, I wrote about cancelling my subscriptions to Lovable, Replit, and Cursor. People asked: why would you drop some of the best AI coding tools out there?

The answer: once you’ve cracked the architecture and structure, the tool itself stops being the bottleneck. The challenge isn’t code generation anymore. It’s intent. It’s maintenance. It’s clarity.

Which is exactly why GitHub’s new Spec Kit caught my eye.

Personal Story When I set out to rebuild an eCOA platform as a challange, I did it in a week. Design, backend, instruments, auto syncing forms all done. A couple iterations later, A detailed spec and I rebuilt the whole thing in a single day.

That was the moment the penny dropped. AI tools could help me get to “working code” faster than ever, but working code wasn’t the problem anymore. What mattered was whether the code matched the intent my requirements, my spec, my design choices.

Once the architecture clicked, the tools felt optional. Which is why I cancelled those subscriptions. Not because Lovable, Replit, or Cursor were bad. They’re excellent. But without a better way to capture and translate intent, they weren’t essential.

That’s the hidden flaw of today’s AI coding. The model isn’t the problem. We are.

We throw in broad prompts.. “add captuer user credentials to my app” and expect magic. What comes back is often correct looking code that’s not what we really meant. The AI isn’t wrong. It’s guessing.

Spec Driven Development: The Fix Spec Kit flips the process. Instead of coding first and documenting later, you specify first, code second. The specification becomes the source of truth — not an afterthought.

GitHub’s Spec Kit formalises this into four gated phases:

1. Specify – Capture the “what” and “why.” The AI generates a living spec with user journeys, acceptance criteria, and “needs clarification” notes.
2. Plan – From spec to architecture. Stack choices, data models, and even rationale are explained. Transparency, not black boxes.
3. Tasks – Break it all into numbered, testable units. No vague dumps.
4. Implement – Incrementally build and verify, people in the loop. Cleaner, safer, reviewable code.

It’s structured, it’s transparent, and it eliminates the guesswork.

The Clinical Trials Parallel This reminds me of the Charter we used in clinical trials development.

At Clario, people thought the Charter was just another document., it was the source of truth. It aligned developers, testers, sponsors, and regulators on exactly what was being built, why, and how it would be verified.

That clarity let us halve delivery timelines and reduce defects dramatically.

Spec Kit is the same idea, repackaged for AI coding. It takes what we already know in pharma that specs and protocols save you from chaos and applies it to software.

Outcome...

• Cleaner, safer builds – Whether it’s trial systems or apps, clarity upfront beats debugging later.
• Trust in automation – AI can only build well if it knows exactly what to build.
• Scalability – Without specs, scale = chaos. With specs, scale = consistency.

Lastly.... I didn’t cancel my subscriptions because I lost faith in AI tools. I cancelled because the tools solved the wrong problem. Spec Kit solves the right one: the gap between intent and implementation. It’s not about guessing better. It’s about specifying better.

CTA So here’s my question: if specs transformed clinical trial delivery, can they now transform AI coding? And have you tried Spec Kit yet or do you think this is another layer of overhead? Let me know your thoughts...

The Clinical & AI industry is evolving in real time, and we’re all figuring it out together. If you need help.. or just want help... Give me a call... :-)

#Innovation, #Leadership, #Entrepreneurship, #CareerDevelopment, #FutureOfWork, #Management, #Creativity

********************************************************************************

Leadership is seeing the future clearly and choosing to build it with integrity.

Stokkan Bray is Founder & CEO of 6ith, a purpose driven company developing eCOA Solutions. He writes about Clinical Trials, AI & Leadership. To learn more, connect on LinkedIn and follow the journey...

https://www.garudax.id/in/stokkan-bray/ and https://open.substack.com/pub/stokkan/