tf Labs 001: Uncanny Valley

Curiosity does not fit neatly into a sprint. It turns up in the half hour between meetings, in the thing someone builds at home and mentions too casually on a Monday, in the question that would take a day to answer properly and so never gets answered at all. We kept noticing how much of that thinking the studio was carrying around unspent, and tf Labs is what we did about it: one full day, every quarter, where nobody touches client work and the only thing anyone owes is a real attempt at a real idea.

The shape is deliberately simple. Pods form in advance and are cross-discipline by design, so a strategist, an engineer and a designer who have never shipped together end up at the same table. The theme stays sealed until nine on the morning, which rules out quietly preparing for weeks and puts everyone at the same standing start. From there it is a full build day, with teams presenting to the room in the late afternoon to a judging panel that includes a guest judge from one of our partners.

What a team owes by five o'clock

Three things, and the first one is the one that keeps everybody honest.

1. A working demonstration. Not slides describing a demonstration, not a video of one that worked earlier: the thing, running, in front of a room that will ask about the bit you were hoping to skip.
2. A pitch that makes the value clear - what this is for, who it serves, and why it is worth the studio's attention beyond the fact that it was fun to build.
3. An implementation roadmap, which is the unglamorous deliverable that separates an idea from a demo and is exactly why we ask for it.

Judging runs on a published rubric rather than a show of hands: eight criteria at ten points each, covering the quality of the insight, originality, business value and strategic fit, feasibility and the path forward, market readiness, the pitch itself, how the team worked together, and how honestly the build answered the theme. Publishing it in advance changes what people build. A rubric that rewards teamwork and a credible roadmap alongside the demo makes it much harder for one heroic engineer to carry a pod to a win, which is the point. There is up to R10,000 in prizes on the day, divided against those scores.

Everyone gets their Claude credits topped up for the day, and how a team spends them turns out to be its own small test. The pods that treat the credits as a shared resource and split the work across people get further than the ones where a single person disappears into a session for three hours while the rest wait to be told what was decided. Learning to work alongside the tooling, in a group, under time pressure, is half of what the day is actually teaching.

001: Uncanny Valley

The first edition ran on 26 March 2026, and the theme revealed at nine was Uncanny Valley - which the brief framed less as a technical challenge than as a standard to be held to.

“Imagine the solution that doesn't feel like a solution. The product someone uses without noticing they're using it, the service so fitted to a real human need that it feels less like an invention and more like something that was always missing.”

It is a harder brief than it first sounds, and you could watch it land. A theme like that quietly disqualifies the demo that impresses by being visibly clever, because anything you notice yourself using has already failed the standard. Several pods spent a chunk of their morning arguing about what invisible even means before writing a line, which felt like lost time at eleven and looked like the right call by four.

The pods, and what they built

[TO COME: the pods, what each one built, and how the judging landed. Send through the team line-ups, a couple of lines on each build and the result, and this section becomes a paragraph per pod with the winner at the end.]

Why the day stays in the calendar

A Labs day costs real delivery time four times a year and we would defend the line item to anyone who asked. Some of what gets built is genuinely useful and some of it is never spoken of again, and both outcomes are fine, because the reliable return is not the prototypes. It is a studio that has practised working with people it does not usually sit near, on a problem nobody has already scoped, with a deadline that does not move. That is a difficult thing to teach in a retro and a straightforward thing to learn in a day.

Edition 002 follows in June, and will deliberately break the pattern. There is no theme to hold steady this time, no shared problem to interpret, just an open brief to prototype an entire product in the day and have something working by the end of it. Where the first edition tested whether people could hold a difficult idea steady under a constraint, 002 will test whether they can finish one without any constraint to lean on at all. Different test, same instinct: put the studio somewhere it does not usually stand, and see what it does there.