How I got here
I started on my grandparents’ ZX Spectrum in my early teens. It drops you straight into BASIC from the main menu, so I wrote little text adventures, saved them to tape and made the family play them. That was where I learned variables, loops and state, without knowing that was what they were called.
After school I went straight into a BTEC in interactive media at Weymouth College, finishing with a distinction, then studied games technology at the University of the West of England. There I wrote the first version of my Lorenz cipher machine simulator, and, in the summer of 2014, my first real game, Escape Sequence.
My first job was at Resort Marketing in my home town, where I built over a dozen iOS and Android apps for clients and learned to work to timelines, specifications and a team. In 2018 I joined Admit One, which makes the software cinemas run on. I built point of sale, ticketing, kiosk and catering software, wrote drivers for receipt printers and pin pads, and helped keep customers running through COVID.
In 2021 I became lead of the core development team. I moved us from single-lock source control to git and agile, reviewed every pull request, and wrote the build manager that packages the whole suite for customers. Since January 2026 I have led the Development department: nine people across core, web, QA and data engineering, with software used by hundreds of cinema customers in eight countries. I still write a lot of the code myself.
In 2026 I also founded Dashcat, which lets people look after their Windows, Mac and Linux machines from the AI apps they already use. It is the project where my interests meet: real machines, real access, and security designed in from the start.
How I work
- Check, don’t trust.In Dashcat every command is signed, names one machine and expires in 60 seconds, and whatever comes back from a machine is treated as data. In the reverse-engineering project a compiler, not a person or a model, decides what counts. I like systems that prove they’re right.
- Write decisions down.Dashcat has more than 60 architecture decision records and the decompilation has 37. When a decision changes, the record says so and why. The reasoning should outlive the conversation it came from.
- Small and dependable.The Dashcat agent has a budget of 10 MB and idles at about 5. Replies are capped in size, and nothing retries silently. Software people rely on shouldn’t surprise them.
- Teams, not shortcuts.I moved the core team from single-lock source control to git and agile, and later brought the web team into git and Docker. Good tools should help a skilled team do more and better work, not make it smaller.
What I’m working towards
The question I keep coming back to is how to give software real access to real systems safely: checked, not trusted. Dashcat is one answer for machines. The injection experiment is a way of measuring how well answers like it hold up, rather than assuming they do.
I would like to spend more of my time on that question, alongside people who study it full time.
Outside work
I build things to understand them. The Lorenz simulator has become my “hello, world”: I rebuild it in every language I learn. I have a long-running strategy game, Pocket Wars, and I am matching Resident Evil 3 back to its source code, byte by byte. Over the years I have built a Nintendo Switch tool used by hundreds of people, a podcast soundboard that fans kept adding to, and an app for controlling a PC from your phone that, looking back, asked the same questions Dashcat does. They are all in the Lab.
The fastest way to reach me is tomtaylorxxi@gmail.com. My CV has the details.