Alongside client work we build our own apps on the same stack. They keep us honest about what works, and the lessons go straight back into client projects.
LIVE · IN PUBLIC BETAHEALTH · PWA · SUPABASE
WaitWell
A companion app for anyone stuck on an NHS waiting list. It turns a long, silent wait into something you can track and prepare for, instead of something that just happens to you.
Daily tracking for pain, mobility, mood and sleep, with trends you can show a clinician
An AI coach that reads your own logs and writes a summary for your appointment
Photograph a hospital letter and get it explained in plain English
Verified NHS guidance links, and installs to your home screen like a native app
Ask a plain question about PIP or Universal Credit and get a plain answer, with the GOV.UK guidance it came from linked underneath it. It points you at the official guidance rather than pretending to be an adviser, and it says so when the guidance does not cover your question.
Twenty three pages of GOV.UK guidance on PIP, Universal Credit and appeals, pulled through the official Content API
Vector search over the guidance, so questions match meaning rather than keywords
Every answer carries its source, and refuses to guess when there is no source
Python and FastAPI behind it, and a progressive web app that installs on any phone
COMING SOON · PHASE 1 IN BUILDRASPBERRY PI 5 · VOICE · OPEN BUILD
Project TARS
A companion robot built in the open, one evening at a time. Phase one has a single goal: give him a voice. A Raspberry Pi 5, a microphone, a speaker and a small display, running a speech stack that listens, thinks and answers out loud with a personality of his own.
Everything after that is earned. Eyes come next, then motors, then the part where he moves under his own power. The whole build is filmed and posted as it happens, mistakes included.
Raspberry Pi 5 with active cooling and the official 27W supply
USB speakerphone for far-field listening and clean playback
Pimoroni Display HAT Mini over the 40-pin header for his face
Speech in, language model in the middle, speech out, with a voice tuned to sound like him
Phase 1 · Voice. He listens, answers and holds a conversation
Phase 2 · Vision. A camera, so he can see who he is talking to
Phase 3 · Motion. Motors, a body, and the first steps
Phase 4 · Out of the workshop, into the world
The build is documented day by day on Instagram, from the first box opening to the first thing he says. No edits that hide the failures.
BCM2712 Quad-core Cortex-A76 at 2.4GHz, the part doing the thinking
LPDDR4X 2GB of system memory
RP1 The input output controller, handling USB, ethernet and GPIO
PMIC Power management, fed by the 27W supply
40-pin GPIO Where the Display HAT Mini sits
Plainly Benefits Live
Claiming Personal Independence Payment or Universal Credit means reading a lot of guidance, usually under stress. Plainly Benefits answers questions about both, in plain English, using only official GOV.UK guidance. Every answer links to the page it came from, so the person can read the original and check it.
Pulls the PIP, Universal Credit and appeals guidance from the GOV.UK Content API.
Splits it into short passages and creates embeddings with an open-source model.
Stores the passages in Postgres with pgvector.
For each question, checks the guidance actually covers it, then writes an answer from the closest passages with numbered links back to GOV.UK.
Provides information from official guidance. It is not benefits advice, and it is not affiliated with the DWP or GOV.UK.
Status
Live and free to use, installs on any phone
Stack
Python, FastAPI, Supabase (Postgres and pgvector), fastembed, Gemini
Sources
23 pages of GOV.UK guidance on PIP, Universal Credit and appeals
For
People making a claim, advice workers, charities and councils
Crypto market research In development
A research desk in the browser for anyone trying to understand crypto markets without twenty tabs open. Live prices, a sentiment gauge, a scenario calculator to test what-ifs, and paper trading to practise without risking money. The Deep Scan tab uses AI to summarise a coin's fundamentals and recent news in plain English.
What's in it
Live prices and market data from CoinGecko
Fear and Greed sentiment meter
Scenario calculator for what-if positions
Paper trading with a virtual balance
AI Deep Scan summaries
For information and education only. Nothing in the app is financial advice or a recommendation to buy or sell anything.
Status
In development
Stack
React, CoinGecko API, LLM APIs
For
People learning to read crypto markets
Reels pipeline In development
An automated pipeline for producing short-form video without a presenter on camera. It researches a topic, writes the script, generates the voiceover and captions over stock footage, then waits for a person to approve before publishing to Instagram and reading the results back into the next round.
The five stages
Research the topic and pick an angle.
Create the script, voiceover, clips and captions.
Human review, with edits or a rejection.
Publish through the Instagram Graph API.
Pull analytics back to inform the next run.
Status
In development
Stack
Python, Instagram Graph API, text-to-speech, video tooling
For
Creators, small brands and agencies
Want something like this for your business?
We build the same kind of product for clients, on the same stack, with you owning the result.