The network
Each one started as a problem I hit while building something else. They share infrastructure, a design language and a rule: say what is finished and what is not.
Newest status check
Finished enough to call live, and publicly available right now at the address shown.
A quiet corner of the web for breathing, resting and doing nothing in particular.
What is different
It removes the machinery that usually keeps you in an app — no streaks, no leaderboards, no feed, no message about the day you missed.
Sixteen free calculators and utilities for UK work, money, documents and everyday admin.
What is different
No account, and the file-handling tools do their work inside your browser, so a document you compress or merge is never uploaded anywhere.
A catalogue of developer utilities — formatters, converters, encoders, generators and testers — that run in the browser, with no sign-up.
What is different
Most tools process input entirely on your own machine, and the catalogue says which is which. The few tools that need the network say where a request goes before anything is sent, rather than after.
Takes a messy PDF, CSV or Excel file and produces a cleaner, more structured copy — repeated headers stripped, ragged tables squared up — while the original stays exactly as it was.
What is different
The file is processed in your own browser and never uploaded. Every change is listed so you can see what was done, anything uncertain is held for your review instead of silently applied, and the original bytes are never touched.
A search engine that mixes live general-web results with a small, independent index I crawl and rank myself, and says plainly which is which.
What is different
Most of what you see when you search is drawn from Brave's general-web results, labelled as such. Alongside those, a smaller set of results can come from my own crawler and ranking, labelled separately. Neither is presented as the other.
A collection of test-design, test-data and automation tools for testers and developers, built so most of them never need to send your input anywhere.
What is different
The great majority of the tools run entirely in your browser and work with the network disabled. A small, clearly marked group needs the network to do its job — checking a live URL's headers, for instance — and those are the only ones that send anything out.
A workspace for modelling cash, revenue, costs, hiring and funding, where every calculated figure — burn, runway, break-even — can be clicked to show the formula, the inputs and the engine version behind it.
What is different
Money is handled as exact whole-unit and exact-fraction arithmetic rather than ordinary floating-point numbers, and every input is classified as an actual, an assumption or a calculation, enforced in the underlying types rather than left to convention. Nothing in the calculation path is produced by an AI model.
Reachable and working today, and still being built. Each one says on its own page exactly what is missing.
A UK-focused way to tailor a CV you already have to a specific job advert, without rewriting it from scratch.
What is different
Every line of a tailored CV has to trace back to a claim you confirmed was true. If a sentence cannot be traced, the document is not delivered.
A place to diagnose an engineering problem by running tests, rather than by guessing from an answer that worked for somebody else in 2019.
What is different
Evidence is attached to a specific revision and a specific environment. Editing a playbook creates a new revision, and the old evidence stays with the text it actually tested instead of silently transferring.
A news reader built around one question: what actually changed while you were away?
What is different
Sources, claims, corroboration, editorial review and revision history are kept as separate signals, so a reader can see how well a story is supported rather than only what it says.
Projects in development, research and experiments are kept separate, because none of them is something you can use.
A pre-alpha, Rust-built research kernel developed and verified entirely inside QEMU, exploring what a personal operating system could look like with explicit privilege boundaries built in from the start.
A Chromium-based desktop browser, built on CEF, with the sandbox on and AI structurally off by default. Phase 1 of a planned four is complete on Windows; nothing has been released.
A research project building a phone operating system in which AI agents may propose actions, but only separate, deterministic code can permit them to happen.
A private, in-development platform for software project governance — lifecycle, traceability, impact analysis and evidence-gated approvals, held to its own standard.
A private, in-development UK toy storefront: catalogue with toy-safety compliance gates, cart, checkout, fulfilment, returns, support and admin with MFA. Not open.
An experiment that put a four-qubit variational circuit at the end of a classical machine-learning pipeline, ran it on real IBM hardware, and measured what that actually bought. The answer was not an accuracy win, and the repository says so in its own first paragraph.
A research question about whether organisations could share tested ERP migration patterns with each other without sharing anything about their customers. There is no name, no domain, no prototype and no code.
An early question about whether posts could become structured, connected, evolving information instead of isolated items that fall down a feed and are gone. Unnamed, undesigned and unbuilt.