Intent-Driven Development · founding axiom

The quality of AI output is determined by the quality of the intent you express.

Scott Feltham — creator of FORGE · 25 years of engineering leadership · CTO, proving it daily

Scott Feltham
the engineer
1×engineer
1,780+commits · 4 months
215+tagged releases
11new services shipped
3.3×ai cost per unit, down

The working proofOne engineer. Twenty repositories. No heroics.

As CTO of a PR-technology company, I am the entire engineering department: a multi-product AI platform — high-volume LLM conversation agents, a data lake stood up from zero, a security-first integration gateway, dashboards, a realtime voice agent — run in production by one person. Eleven new services born in four months.

Not by working nights. By working differently: agentic AI systems carry the execution, and my job is expressing intent precisely enough to trust them — then verifying they delivered it. The same discipline re-engineered the platform's AI economics on evidence: cost per message down 3.3× while volume doubled. Trunk-based delivery, infrastructure entirely as code, agentic monitors on cost and health. It is Intent-Driven Development, practiced at its logical limit.

most AI failures are intent failures

— so make intent the artifact

The methodologyFORGE — Intent-Driven Development

Most failed AI-assisted development isn't a model-capability problem — it's an intent-clarity problem. FORGE makes intent expression a first-class engineering activity, with phases, deliverables, and quality gates. You don't earn the right to generate code until you can describe what done looks like.

F

Focus

Capture intent: what, why, and success in testable terms.

O

Orchestrate

Decompose intent into work AI and humans can carry.

R

Refine

Make it verifiable: acceptance criteria, given-when-then.

G

Generate

Now the code — by agent, pair, or hand — traced to intent.

E

Evaluate

Prove outcomes match intent, with evidence on record.

Intent

The precise articulation of what you're building and why it matters — sharp enough for an agent with no intuition, only instructions.

Outcome

Intent without a measurable outcome is a wish. Every line of code traces to an outcome; every outcome traces to an intent.

Accountability

The verifiable chain from intent through outcome to evidence — what makes autonomous development trustworthy.

documentation · live

The full methodology is documented and free to read: the FORGE Framework docs — getting started, the five phases in practice, agent patterns, and tool integrations. Intent-Driven Development, the book, follows. docs live · book v1.0 complete

The recordTwenty-five years, measured in outcomes

  • The long apprenticeship · defense systems → gov digital → mobile → SaaS

    General Dynamics, freelance, Foster Findlay, Orange Bus (HMRC Tax-Free Childcare), Nutshell, Condeco.

  • Senior Engineer & Ops Manager · Moltin / Elastic Path

    Distributed SaaS operations held steady through acquisition.

    zero downtime through the transition

  • Head of Engineering · mkodo

    First SaaS product for regulated gaming markets, concept to production.

    0 → 1 in 8 months

  • Senior Technical Architect · Builder.ai

    Founded and led the Architecture Guild; technical lead on GCC-region pre-sales.

    8 engineers mentored to senior roles

  • VP of Engineering · Kahunas

    Led a 20+ person organization through an AI-first delivery transformation.

    2× developer productivity · 99.9% availability during scale-up

  • Founder · NeoForge Consulting & Neurocademy

    Outcome-owned AI adoption programs for technology leaders; an AI lesson-planning platform for neurodivergent learners shipped 0→1.

    concept → working product in 15 days of a 30-day hackathon

  • Chief Technology Officer · Pathos Communications

    Sole engineer for a multi-product production AI platform, operated through agentic systems — where the story so far was heading.

    1,780+ commits · 215+ releases · 11 new services · AI cost per unit ÷3.3 at doubled volume

Plainly statedQuestions, answered

What is Intent-Driven Development?
Intent-Driven Development (IDD) is a software-development philosophy for the AI era, created by Scott Feltham. Its founding axiom: the quality of AI output is determined by the quality of the intent you express. IDD rests on three pillars — Intent, Outcome, and Accountability — and makes intent expression a first-class engineering activity with its own phases, deliverables, and quality gates.
What is the FORGE Framework?
FORGE is the delivery methodology that implements IDD, in five phases: Focus (capture intent), Orchestrate (decompose it), Refine (make it verifiable), Generate (produce code traced to intent), and Evaluate (prove outcomes match intent, with evidence). It works with any AI tool, language, or team structure.
Who is Scott Feltham?
An engineering leader with 25 years across defense systems, government digital services, SaaS, and AI platforms — VP of Engineering, Head of Engineering, and currently Chief Technology Officer. Creator of FORGE and Intent-Driven Development, and author of the forthcoming book of the same name.
Can one engineer really run a production platform through AI?
Yes — with the right discipline. As CTO, Scott operates a multi-product production AI platform alone: 1,780+ commits and 215+ tagged releases across 20 repositories in four months, 11 new services shipped, while cutting AI cost per unit 3.3× at doubled volume. The method is the point: agentic systems carry execution; the engineer's job is precise intent and rigorous verification.
Where can I read more?
The full FORGE documentation is free to read on this site; the book, Intent-Driven Development, follows. Code and frameworks live at github.com/scottfeltham.

05 · correspondence

Thinking about what AI actually changes for engineering? So am I — out loud.

scott@neoforge.co The methodology LinkedIn GitHub

Writing, speaking, and building in public on Intent-Driven Development.