Time taken: 1h
an invitation to my LOTR-themed summer party
Shipping in <1h takes extreme discipline. A single HTML file can produce really strong results. You can do a lot in 1h if you believe in yourself.
Time taken: 6h
An interactive explainer on the Kelly Criterion and gambler's ruin.
Even a simple website explaining a fun mathematical concept as a bunch of interactive statistical models can end up taking 6 hours. a.k.a. EVERYTHING TAKES LONGER THAN YOU THINK IT WILL.
Time taken: <2h
onboarded my first users! let's match some co-founders!
Onboarding your first user is terrifying. Going back to an old build feels super stale, even though doing the work to debug is, in practice, incredibly quick. I should balance building new things with the thrill of engaging with real customers on old things.
Time taken: 3h
Downloadable tinder card generator for your worst possible co-founder
Everything takes 3x longer than you think it will. This includes if you think it will take 20 minutes & plan for an hour (it will still take 3 hours). html2canvas is deceptively fiddly. Even if you don't feel like it, a silly quick build can still be super fun and can still teach you something. The more complete the vision, the quicker the initial prompt. Also, the quicker all subsequent fixes are. Don't be afraid of vision.
Time taken: 11.5h
Finding work-friendly cafés...
Doing one simple thing *really well* takes longer than you think. Scraping the internet for data is tricky and costs $$$. Juggling 4 APIs is a massive pain. It's not straightforward to find a user's location either.
Time taken: 4.5h
Almanac to connect with the sky, the seasons, and the natural world
Give inspiration time to find you - it will. The twinkling star bug took longer to fix than the entire settings page. I still don't think the wheel of the year is quite right, but the perfect is the enemy of the good.
Time taken: 3h
A physical LED tracker (built with an Arduino Uno) to stop me forgetting my ADHD meds.
Hardware is way easier than I expected. Jumper wire connector housings are designed for prototyping, not for fitting neatly inside a box (this caused my only real problem of the day). A Jo Malone box is almost exactly the right size for an Arduino Uno + half-size breadboard. The hardware component took 1h15m; the craft component took 1h30m (i.e. making the box took longer than making the thing). This was the fastest build of the year so far.
Time taken: 6.5h
A tool to make responding to emails 10x faster
You really need to understand your own workflow to build yourself an agent. If you want to avoid spending extra $$$ on API credits you can bypass this by making a private product and using MCP instead. Perhaps gamifying every aspect of my life is the key to success.
Time taken: 5.5h
Chrome extension that detects and highlights ads and sponsored content across major websites
Everything takes longer than you think (even when you already know everything takes longer than you think). Every website needed a completely different detection strategy. Facebook actively hides its ads from browser extensions (obfuscated DOM, rotating class names, character-by-character text splitting, hover-gated elements). Collect bug reports using Google Forms + Sheets.
Time taken: 7.5h
Upload a choral score PDF, get interactive playback with per-voice muting so you can practise your part.
This idea could NOT be achieved in my originally planned 1h, let alone 7.5h. Optical music recognition is hard, slow, and painful. My backend needs Java 17, Tesseract, and Audiveris installed, so couldn't deploy it without paying for another service; live site is demo-only. Audiveris is a bit shit anyway. Building an alternative to Audiveris is 'a startup, not a weekly project'. It's 3:30am and I'm annoyed.
Time taken: 8.5h
Games to play on Valentine's Day for when your date (inevitably) gets boring
Even simple builds with minimal deployment errors take time. Prompting an LLM (via API) is harder than it looks, especially if you don’t want it to be repetitive. It also costs money each time the user prompts it (even though it doesn’t cost much). ALWAYS LEAVE EXTRA TIME TO DEPLOY. Especially if you’re calling a bunch of APIs. It's a bit of a faff to track stats (I had to sign up to a whole new service called Posthog) but should do this for all builds really.
Time taken: 24h
A tamagotchi-like game for Supercell GameDev Hackathon
Team of 2 with Brittany Gibbons (StoneFruitStudios). 3D games are built differently to 2D games. Blender is super cool. 3D assets are complex & take forever to create. I enjoy animation. It’s very different building in a team vs. building alone. Design documents are useful (& take some getting used to). Claude Code > other LLMs. Don’t push to the wrong branch on GitHub. Game assets can be huge (common problem, makes deployment tricky).
Time taken: 15.5h
Built a co-founder matching platform for The Tech Bros community
Have been wanting to build this for over a year. Built the basic functionality in 2.5h. Then had to fix UI / UX. Fixed bugs. Spent ~6 hours on setting up SMS verification (+ getting the interface to reflect real phone numbers with real area codes). Overall, this felt like a breeze compared to 1) previous weeks and 2) what I was expecting this project to feel like. Learnt a lot about how “real” products work.
Time taken: 30h
Created a “let’s invade Greenland” game inspired by recent events
Civ wasn't built in a day. This is still unfinished. Games are hard and complex. Most of your time will be spent debugging. You'll have to build games within games just to get the rules of the game to work. It's hard to get the AI opponent to act logically. Despite all this, games are still insanely fun and I wish I had longer to make this one everything it could be.
Time taken: 23h
Redesigned The Tech Bros website and rebuilt it from scratch
Designing and populating the website felt tedious. The backend / server stuff was more fun this time around. I created 3 separate forms and integrated them with Slack, Airtable, and EmailOctopus so I get a notification when people register for the community AND all their info appears on Airtable directly. Time-boxed sprints are ideal for front end decision fatigue.
Time taken: 30.5 hours
Updated my singing website and created this one in preparation for weekly projects
The server stuff was a bitch. I felt inspired building both websites. Populating a website with content takes many, many hours. Everything takes longer than you think. Don't be afraid to ask ChatGPT to create a Cursor prompt (they work well as a team). Send ChatGPT screenshots when troubleshooting. I'm a perfectionist & will need to work against this if I want to get anything shipped this year at all.
This is my weekly coding diary for 2026.
I come from a mathematics background, so for a long time I’ve assumed that one day I’d be able to do the coding thing properly. But assumption isn’t the same as fluency - and over time, engineering started to feel like a black box I didn’t fully understand.
I want to change that.
The goal here is simple: to get to the point where building things feels easy, intuitive, and genuinely fun. To stop treating engineering as something abstract, and start treating it as a craft I can use fluently.
So for the whole of 2026, I’m building something every single week.
Some weeks will take a couple hours; some multiple days. Some weeks will be scrapped; some will be ambitious. But the key is that - whether a tool, a site, an experiment, a prototype, or a half-formed idea - each week, I have to ship. That rule exists for a reason. Perfectionism is a fantastic way to never finish anything, so this project is deliberately structured to prioritise momentum over polish.
Alongside the builds, I’ll share what I can about the process: this is as much a record for my future self as it is something I’m sharing publicly.
Follow along for the journey - I can’t wait to see where it takes me!
My name is Dr. Milette Gillow. I’m a mathematician and opera singer by training (University of Leeds, University of Copenhagen, Royal Academy of Music). I’ve worked as a VC, university lecturer, and professional singer among other things, and I now run The Tech Bros: the world’s first all-female, all-technical accelerator. I’ve organised 12 hackathons over the past year with The Tech Bros, but haven’t built as much myself: so my aim for 2026 is to spend a lot more time developing. I love creative work, so I think I’ll enjoy it!