About
Hi, I'm Manzier.
I like taking a messy real-time feed, turning it into outcome data, and seeing what — if anything — actually predicts something. Usually less than I hoped: three "obviously good" signals got dropped from RugScore because the data said they didn't separate, after I'd weighted them at 45 points in the first version. The graph beats my intuition more often than I'd like.
By day I run the technology infrastructure for a public school district in Illinois — networks, identity, phones, and the servers nobody notices until they stop. Nights and weekends, I build things with data. Thirty years of keeping other people's systems up is probably where the next habit comes from.
Being boring on purpose. Receive-only payments. Guards that fail closed. Backups that were restored once, for real, before anyone trusted them. A watchdog, a self-recalibration timer, a nightly diary — because the product is the uptime too.
Measure firstLookahead-free calibration on real outcomes, or it isn't a feature.
Boring on purposeReceive-only payments, fail-closed guards, backups restored before trusted.
Keep it runningSelf-recalibration, watchdogs, a nightly diary. Uptime is part of the product.