I TURN CHAOS INTOFUNCTIONAL BEAUTY

Here's what precision software engineering
with human-centric design looks like:

CanvasPro

2019 - Present

A screen management and media platform I built from the ground up with a two-person team. Large-scale screen installations used to require bespoke software every time. CanvasPro turned that into a unified platform for orchestrating content, live video feeds, real-time graphics, and lighting across complex multi-screen environments.

The platform encompasses multiple software products: CanvasPro One, a media server solution integrated with the Ventuz real-time graphics engine for broadcast-grade venues; Orb, a lightweight, efficient player written in Rust; and CanvasPro Cloud for centralised scheduling and remote device management.

I lead product vision, architecture, and development. CanvasPro is a live product with over five years of continuous development behind it, shaped by a process we've constantly reflected on and adapted as technology and scale have evolved over time. Today, it powers Piccadilly Lights, Ascot Racecourse, and Brighton & Hove Albion.

Wed by Wedd

A site for a North London wedding celebrant, designed and built end to end. Vue, prerendered to static HTML, served entirely from Cloudflare. Enquiries go through a Cloudflare Worker that sends email natively, so there are no third-party services to run.

wedbywedd.com

Bliska Mama

An education platform for new parents. Webinars, e-books, and user accounts. Built with a custom CMS, authentication, and payment integration.

bliskamama.pl

Here are my thoughts:

08

Language is not an interface

Every generation of computing tries to turn human speech into a steering wheel. Generative AI made intent computable, but a chat box is still an engineering shortcut. Why talking to software is usually the slowest way to work, what breaks when you banish visual controls, and where the real paradigm lands.

Read
07

Code is for compilers

A Harvard study on 700,000 developers confirmed what every lead already knew: AI wrote 30% more code, and shipped zero extra features. Generating syntax was never the bottleneck. Software works only when someone gives a damn about the person on either side of the screen.

Read
06

You can't skip the beginner work

Tech companies promise AI will take away all the boring chores. But the boring chores are how beginners learn to become experts. What happens when the easy work disappears, and why real experience cannot be automated.

Read
05

Playing someone else's piano

When OpenAI showcased Astra building 3D scenes in Blender via MCP, commentators declared 3D artists finished. Anyone working in production saw the catch immediately. Why the demo was marketing theatre, and what it reveals about the limits of generic AI.

Read
04

The compiler reviews what you won't

The case for Python and JavaScript was always developer time. Agents made typing free, which quietly removed the argument and left a stricter language holding something more valuable.

Read
03

Contain it, don't review it

If your plan for agentic coding is to read every line, you have traded a speed gain for a queue with you in it. The work that scales is building the walls, not standing in front of them.

Read
02

An assistant, not a mind

The machine that writes half my code is not on a path to waking up, and that was the design brief. Why the replacement question never made sense, and what near-free development does to off-the-shelf software.

Read
01

Build first, ask questions later

The economic argument for waterfall was always that being wrong in code costs 100x being wrong on paper. Agentic coding flattens that curve. What replaces the premise, and what gets harder instead.

Read

A little more about Lee Brown:

I like making complicated things feel simple.

I'm a developer who ended up in the world of giant screens and live events. Over seven years I've gone from writing my first Vue component to architecting the platform behind Piccadilly Lights.

I like building things that have to work. No retries, no "refresh the page", no excuses. If 50,000 people are watching, it just has to be right.

The code is usually the easy part. What I enjoy is the design: sitting with a difficult problem until the right shape appears, then choosing between the proven pattern and the new technology that might do it better. When it works, nobody notices. That is the point.

Digital Media
Technologies

2019 - Present

Where I work. A company that designs, builds, and runs advanced media systems for physical spaces. I lead development and provide technical consultancy across the business.

I also run R&D, including early AI-powered interactivity for out-of-home campaigns, back in 2019 before it was a buzzword.

RoleLead Developer
Since2019
FocusSystems design, real-time AV, cloud platforms, AI interactivity
EducationUniversity of Hull, Computer Science
Stack
Languages

RustProduction, where performance matters

PythonPrototypes, and production via Nuitka where it doesn't

TypeScriptFrontend, usually Vue or Nuxt

Cloud

AWSRDS, DynamoDB, Cognito, MQTT

CloudflareWorkers, D1, Pages

TerraformInfrastructure as code

Real-time

NDI, SDI, DMX, Ventuz, Unreal Engine, WebGL