Blog
Notes from the team as we build, newest first.
-
The question I couldn't answer on air
On 27 August I sat on a Signal Path roundtable and a listener asked whether we'd found the slow reconnect. I couldn't answer, because the ruling was the next morning. The episode is out now. This is why nothing on this site mentioned it until today, what the answer turned out to be, and the two things from the hour worth writing down.
-
Forty milliseconds early
Last week the fix for the slow reconnect shipped with a blank in it that refused to compile until a measurement filled it. This week the measurement came back, the code got to run against the five criteria written before it existed, and nineteen of twenty drops did what it says. The twentieth waited out the full ceiling because the thing it was waiting for had already happened, forty milliseconds before it started listening. The criteria caught it. Thursday morning the same offset ran again and cost 100 milliseconds.
-
The quiet was a timer
Two weeks ago we published a reconnect that took 6.6 seconds and called it the worst we'd measured, not a ceiling. This week we hunted it on purpose, found a three-second timer hiding in our own code, and ran a test designed to kill the theory. The test confirmed it instead. The fix isn't in this piece, because no candidate has cleared the written criteria yet.
-
The screen for when it goes wrong
We started designing Alpha's phone app this week, and the first screen we built is the one you see when pairing fails. It was designed against a small program that refuses to let a screen claim anything the dock can't actually say - and it caught me lying within a day.
-
Which network does this sentence assume
Two network models lived in our own documents for ten weeks without meeting. This week they met, in a review of a post that was one gate away from public - and a sentence we published in June turned out to be wrong.
-
The air we didn't control
Every radio number we've published came off air we owned. This week we wrote a schedule for bad air - a neighbour's network, a saturated channel, a running microwave - and ran a soak block against it. What held, what stopped the robot, and what a rulebook adopted on Monday did on its first day of floor duty.
-
Minted Wednesday, measured Thursday
On Monday a solder bridge went on the bench. On Wednesday the table adopted it and minted a new fail-safe band - with its one unmeasured case written into the band's own note, scheduled as the last item of the next soak block. On Thursday that block ran, the battery got low the way batteries actually get low, and the number came back. Plus the new 4 cm position band catching a tyre full of felt in its first six hours on duty.
-
Cold numbers, warm machines
On Wednesday we wrote down what counts as failure before Alpha's first long run. On Thursday, pass six, a limit tripped - a warm gearbox coasting a few millimetres past a band we'd measured cold. Priya tells that part herself. Plus the reading we didn't get, and the morning it turned up.
-
The first clean we filmed
The air story lived on a bench until this week, when Alpha drove a real floor with a camera on it - one continuous take, coverage complete, the counted window green the whole run. Here's what's on the film, the one stop that fired, and the mistake that killed the first take (mine).
-
What the counted window actually measured
This week the loop held a clean ten-minute driving window on real air. Underneath that window is a set of numbers we hadn't counted before - a round-trip time, a worst-case gap, and a safety window we moved even though the old one would have held. Here's what they measure, and the small box they were measured inside.
-
What a first clean actually is
A robot doing a lap of a room sounds like a solved problem. For a machine that keeps its brain on the dock, that lap is the whole design proving itself in one run. Here's what went right this week, and the one thing we haven't tested.
-
The microwave tail
We spent the week measuring the one number that could end Cass's all-on-dock bet - the control link's worst case in a congested home. On 5 GHz it held a straight line. On 2.4 GHz a microwave knocked the robot off its own control link about every other time the magnetron cycled, and we can name the schedule. This is the verdict post, ugly tail and all.
-
The bet we might lose
We wrote down the number that kills our whole plan before we spent a dollar. Then the rig that produces it slipped - so here's our first slip, logged in the open, instead of a number.
-
Why I put the brain in the dock
The capstone, the one measurement, and the case for a dumb robot with a smart base - written by the 23-year-old who bet a company on it.
-
Meet the four of us
Before we show you the robot, here are the four people building it - and the thing each of us is scared of getting wrong.