How I use AI in development.
The tools are only half of it. The other half is the MCP layer that gives them the design file, the repo, the database, and the ticket — plus a review habit that never goes away.
How AI fits into the workflow
I use AI to move through the boring parts faster — scaffolding, refactors, copy, tests — then read every line and own what ships. The difference is the MCP layer: the tools can see the design file, the repo, the database, and the ticket, not just the chat window.
Daily drivers
- Cursor
- Windsurf
- GitHub Copilot
- Lovable
- Codex
For research, planning & content
- Claude
- Gemini
Tools change every few months. The standard doesn’t: every change gets read, tested, and reviewed before it reaches a user.
Project note / Prophic AI
Making notifications useful
The notification centre brings filtering, pagination and translated messages together. Each notification routes to the relevant workflow, so users can act on it in context.
Read the case study →MCP connections
The project context behind each change.
4 integrations in the everyday workflow, so the assistant is working with the real project — not a summary of it.
Figma
01Design-to-code handoff
Reads frames, tokens, and component structure straight into the build.
GitHub
02Repo, PR & issue management
Keeps branches, reviews, and issue context inside the same workflow.
PostgreSQL
03Schema & query access
Inspects tables and tests queries without leaving the editor.
Slack / Jira
04Task & team sync
Pulls task context and keeps delivery updates close to the code.
Where AI fits
01
Explore
Read unfamiliar code, map a feature, and compare approaches before writing a line.
02
Build
Scaffold components, clear repetitive refactors, write tests, and handle translation passes.
03
Review
Catch edge cases, explain diffs, and tighten the pull request before a human sees it.
Let’s discuss your next project.
Have a product, a role, or a messy front-end that needs a steady hand? I’m one message away.
Start a conversation