Migrating Xamarin.Forms to .NET MAUI
What actually breaks, what the tooling handles for you, and how to estimate one of these honestly.
Insights
What we're working on and thinking about
We spend real time on research and development, partly to stay current, mostly because it is the enjoyable part of the job. This is where we write some of it down.
What we'll be writing about
The topics currently on our own research list, grouped the way we think about them: a fair preview of what will appear here first.
Mobile & .NET MAUI
Cross-platform apps, and the awkward parts
Moving off Xamarin.Forms, picking a Blazor render mode, keeping data in sync when the network is not there. The questions that decide whether a cross-platform app is pleasant to live with, written up from real migrations rather than in the abstract.
Blazor render modes in practice
Server, WebAssembly, auto: which to pick, and the hosting and latency consequences of each.
Offline-first sync for mobile
Conflict resolution strategies that do not quietly lose a user's work, and how to test them.
Building for constrained networks
Designing apps for South African realities: intermittent connectivity, expensive data, and load-shedding.
A holographic display built from a phone and some glass
The Pepper's Ghost illusion is over a century old. We have rebuilt it with .NET MAUI and a phone's own sensors, turning a stage trick into a genuinely interesting rendering problem.
IoT & the edge
When the computer is smaller than the problem
Provisioning devices before the first thousand ship, running .NET on boards that cost less than lunch, and securing a microcontroller end to end. Notes from putting hardware in places it has to keep working unattended.
Azure IoT device provisioning at scale
Certificates, identity and the update path, designed before the first thousand devices ship rather than after.
Running .NET on hardware smaller than a credit card
We have run C# and .NET IoT on microcontroller-class boards that cost less than lunch. The line between a toy project and a genuinely useful skill is thinner than it looks.
Smart composting: sensors for the bin, not just the bed
Yield is not only about growing more, it is also about wasting less. We have added moisture and temperature sensing to a composting setup to see what that actually looks like in practice.
Locking down a microcontroller with Azure Sphere
We have built connected microcontroller devices secured end to end with Azure Sphere: hardware root of trust, OS and cloud security service treated as one system, not three.
Applied AI & vision
Models that have to earn their place
Retrieval that survives real data, inference with no GPU to lean on, and a camera plus a model doing something genuinely useful. The interesting decisions are in encoding, cost and fit, not the demo.
RAG that survives contact with real data
Chunking, embedding and retrieval quality behind the LLM: the encoding decisions that decide whether an AI feature is useful or merely impressive.
Squeezing a useful model onto a Raspberry Pi
We have run edge inference with no cloud GPU or NPU to lean on: what actually fits, what has to be cut, and where the compromises show up first.
An AI-powered microscope, and what it is actually for
We have paired a Raspberry Pi, a camera and cloud vision AI, encoding each frame for a model to reason over, into a low-cost digital microscope. The interesting part was never the demo, it was what it makes possible for someone who could not otherwise access one.
Building AI features into a MAUI app from day one
There is a real difference between a mobile app with an LLM bolted on and one designed around it from the start, agents, MCP tool-calling and Skills included. We have built cross-platform .NET MAUI apps the second way.
Want us to cover something?
If there is a question you keep running into, tell us. It is a good candidate for the first post.