n.V6-2.06 | What We Want From the Machine
01| This is the want list. Same shape each time: what we want, what the field says it takes, what we give up. The technical answer to each one belongs to somebody else and I will not be guessing at it.
02| Robust, to the point of dullness. The platform does not break in front of a group. A Maveriq standing in a room with fourteen people she persuaded to show up cannot have the thing fail on her — that failure is not technical, it is social, and it costs her standing that took months to build.
03| What it takes: redundancy everywhere, graceful failure of parts rather than collapse of the whole, and a release culture that treats stability as more important than the calendar. What we sacrifice: speed, twice over. Robust systems are slower at the edges, because every safety net has a cost in milliseconds. And we ship features more slowly, because everything gets tested harder than it would elsewhere. We take both losses. A missing feature disappoints. A crash in public humiliates.
04| Idiot-proof. I hate the term and I cannot find a better one. It means a person who has never used software like this, on a bad phone, under stress, gets through the task without a manual and without asking anyone for help.
05| What it takes: fewer options, obvious defaults, and forgiving mistakes rather than punishing them. What we sacrifice: power at the top. Our strongest Maveriqs will hit a ceiling and find the tools childish. That is a genuine loss and it will cost us some good people's patience. It is still the right trade, because we survive on what the average member delivers, not on what our stars manage. Depth comes later, for people who have earned their way to needing it — never at the entrance.
06| Available. It works when she opens it. Not most of the time. When she opens it.
07| What it takes: monitoring that catches problems before users do, someone actually on call, and maintenance scheduled when the barrio is asleep rather than when it is working. What we sacrifice: it makes us slower and more conservative than a startup, and it costs real money on nights when nothing happens. Downtime destroys trust at a rate features cannot rebuild it.
08| Secure, because our users cannot absorb a breach. A middle-class person whose data leaks has a bad month, some phone calls, a replaced card. Our users have none of that cushion. In some barrios, the wrong information in the wrong hands is not an inconvenience, it is a danger.
09| What it takes: encryption as standard, collecting less than we are tempted to, tight access control, and treating it as continuous work rather than an audit we pass once. What we sacrifice: convenience and insight. Every additional login step is friction on people who have little patience for friction. Collecting less data means knowing less, which means a worse product in measurable ways. We have already committed elsewhere that we do not track people and do not sell data — this is the same commitment paying its bill. Worse personalisation is the price and I will pay it without argument.
10| Not owned by one supplier. Especially on AI. The whole promise of AI here is that it eventually offloads a great deal of human work and lets us scale at a cost we could not otherwise reach. That promise is worth very little if a single provider can change terms and take our margin, our capability, or our existence with them.
11| What it takes: building so that providers can be swapped rather than woven in, which is more work and less elegant. What we sacrifice: depth and speed. The organisations that go all-in on one provider get better results sooner. We will be a step behind them in capability and a step ahead of them in survivability. For an organisation whose entire claim is permanence, that is not a close call.
12| Flexible enough to change. New insight has to be able to reach the product, or the whole reconnaissance idea is theatre.
13| What it takes: modular construction and the discipline to keep it that way under deadline. What we sacrifice: stability, which is the direct opposite of the first want on this list. A platform easy to change is a platform easy to break, and the tension between those two runs through everything here. Where they collide, robustness wins and the change waits.
