The second product is found by operating the first

What operating a product tells you

The useful thing about running software in production is not the feedback form. It is that a working product puts you in a room with people who have an adjacent problem you were not looking for, and keeps you there long enough to notice.

Windline forecasts wind along a cycling route — it reads the bearing of every stretch, works out when the group reaches each one, and reports headwind rather than a compass direction. Operating it surfaced three facts in order: a large share of those riders also run, the running population is considerably larger than the cycling one, and running has nothing like the same relationship with wind.

A roadmap is a guess about what matters. An operating product is a standing instrument for finding out.

Wind is the cyclist’s enemy. Heat is the runner’s.

A cyclist at 25 km/h generates a constant stream of moving air. That airflow does two jobs: it resists you, which is why wind direction decides a ride, and it cools you, which is why heat is survivable at speed.

A runner has neither. At running pace there is almost no cooling airflow, so the body loses heat by evaporation and little else — and whether evaporation works depends on how much moisture the air can still take. That is why dew point, not temperature, is the number that ruins a race.

So the same day is two different days. For a rider it is a question of which direction the wind is coming from. For a runner it is a question of whether the air can still take sweat, and the honest answer changes hour by hour as the day warms up.

What nothing else was doing

The gap was specific rather than general. Weather for runners exists; weather timed to a runner does not.

• Race tools cover famous courses — useful if you are running one of perhaps forty events in the world, useless for the route outside your door.
• Pace calculators take a single temperature for an entire run, which is a reasonable approximation for forty minutes and nonsense across four hours of a warming morning.
• Route-weather tools are built around cycling speeds, so they resolve in kilometres per hour and think the interesting variable is wind.

None of them answers the actual question, which is what it will feel like at your pace at kilometre 31 at the hour you get there.

Retuning an engine rather than building one

What made this a fortnight’s work rather than a year’s is that the hard part was already solved. Walking a route in time, resolving conditions to a point and an hour, blending between model readings — that machinery does not care whether the thing moving along the line has two wheels or two legs.

What it cared about was speed, and the assumptions underneath it. Pace is now handled in minutes and seconds per kilometre or mile, in five-second steps, from 2:30 to 15:00. That upper bound is deliberate and it is the detail that says who the product is for: at 15:00/km a run-walker finishing Comrades in twelve hours is timed as precisely as somebody at the front, which is not true of tools that assume everyone is fast.

The output changed too. Rather than a number to interpret, RunFast states the weather cost — how many minutes the conditions will take off you. A forecast becomes a pacing decision.

Why the followers matter more than the runners

The part that will probably decide whether RunFast is bigger than Windline has nothing to do with forecasting.

Runners race in mass events. Comrades alone has twenty thousand runners, and behind each of them is a small group of people at home who want to know where they are. A runner sets a goal, taps go live, and shares one link: followers see position, pace and projected finish against that goal on a map, with no app and no account.

Those followers can send a tap on the back, and the runner feels it mid-race. Which means every follower is a non-runner meeting the product on the day it matters most to somebody they care about — and a good share of them run, or will. The runner shares a link; the followers become the next runners and the next clubs. Cycling has no equivalent of this, because cycling has no Comrades.

Sisters, not a parent and a child

It would be easy, and wrong, to describe RunFast as Windline for runners. The products share a forecast engine, a Strava connection and a live-tracking server, and neither is built on top of the other. RunFast is not a Windline feature and not a product by Windline; the engine is common infrastructure and both are equal consumers of it.

That arrangement is also the test. Shared infrastructure that genuinely generalises can carry a second product without either one bending around the other, and one engine means the numbers are identical wherever they appear — a correction to either improves both. If it had needed forking, the engine was never general; it would just have been Windline with a different label.

RunFast launches for the 2027 season — Two Oceans in April, Comrades in June — and waits on no race organiser to list a course. It is free, with ProLend as headline sponsor in South Africa, and it carries its own December challenge, the RunFast 200: 200 km between 16 and 31 December, counted from Strava.