Field note
AI-native software still needs rigour
What old and new codebases taught me about building with AI without surrendering control or quality.
Read field noteAbout · Dave Hudson
I build software, shape the work and stay close to the product.
A Contract AI Product Engineer with a frontend bias, full-stack depth and more than twenty years of experience turning uncertain product ideas into working software.
Where I sit
The useful overlap between product thinking, interface engineering and modern AI-assisted delivery.
- 01
Frontend by instinct.
React, TypeScript and Tailwind are where I move fastest. I care about composition, interaction quality and the architecture beneath both.
- 02
Full-stack in practice.
I use Node.js and Convex when the product needs a complete vertical slice, especially in small teams and greenfield work.
- 03
AI-native.
Claude Code and Codex speed up the work. Product context, narrow scope, tests, visible checks and human-controlled MCP workflows keep it honest.
HOW I WORK WITH AI
AI is core to how I engineer.
I use Claude Code and Codex throughout the work. Product context and agreed scope guide each task. Tests, visible checks and human approval decide whether it ships.
- Planning and design tools connected through MCP
- Agents get context from the tools the team already uses
- Tests, QA and verification before approval
DELIVERY WORKFLOW
- 01IntentHUMAN
- 02ContextTOOLS
- 03BuildAGENT
- 04EvidenceCHECKS
- 05ApproveHUMAN
CHECKS → BUILDAPPROVAL → CONTEXT
WORKING SETUP
MCP-connected planning and design Context from the team's existing tools Tests, QA and human approval
DELIVERY WORKFLOW
Checks and approval can send the slice back.
- 01IntentHUMAN
- 02ContextTOOLS
- 03BuildAGENT
- 04EvidenceCHECKS
- 05ApproveHUMAN
Failed checks return the work to Build. Human redirects return the work to Context.
Writing from the work
The thinking behind the workflow.
What changes when agents write more of the code, and what engineering discipline still has to do.
View all writingField note
AI is making me rethink software delivery
How AI has changed the way I write, design, test and build software, and what that means for ownership.
Read field noteSelected timeline
More than twenty years, condensed.
The through-line is consistent: understand the product, make the architecture legible and help a small team ship it.
Founded Applification Ltd
Independent consulting, product delivery and software engineering.
Mobile product teams
React Native, Node.js APIs and delivery leadership across travel and public-sector products.
Secure government work
Security-cleared frontend delivery for Cabinet Office and other government agencies.
Clinician platforms
Pando and Peppy: reliable admin surfaces, component systems and testing practices.
Greenfield contract builds
Sole frontend ownership for Eruptiv, building the recruitment product in three months and taking it to production after four.
Rebuilt the frontend in Next.js, then co-built Agentic Chat for threat analysts with AI SDK UI, MCP and Databricks tools.
AI product R&D at Applification
NowBuilding StoryLoops, Contexture and Voiced, with each product testing a different use for agentic workflows.
The work I am looking for
Contract roles only. Remote by default; North East hybrid considered.
Best fit
A small team with a product to shape.
- Greenfield or architectural reset
- React, TypeScript and Tailwind
- AI product interfaces and agent workflows
- Direct access to product decisions
Deliberate focus
AI product engineering, not data science consultancy.
I build LLM-enabled products, coding-agent loops, MCP integrations and the interfaces that make them usable. I focus on hands-on web product engineering, not specialist Python, RAG or big-data consultancy. I take on contract work rather than permanent roles.
Credentials
Certified Scrum Master · Certified Product Owner · Government security clearance held for relevant engagements