
software should be able to adapt to you


software should be able to adapt to you

you cant ask most ai decisions why they happened. a person you can at least pull into a meeting. idk how institutions are supposed to work when more and more of the calls come from something that cant explain itself

Here's an example of how Lodge can run a paper mill's power plan. It forecasts real-time price spikes, shifts the heavy load around them, and nothing reaches the mill until a person approves it.

Every technology before this was wielded by intelligence (e.g. a printing press, a rifle, a database). This one is the intelligence. That changes who holds power in a way we have no precedent for.


The Raiders won 3 games last year and are 2-0 now, but they play like the 16th-best team in football. Real or fake for the Raiders, Bengals, Eagles and Seahawks. Every Week 3 call was locked before kickoff, and we grade them all next week in public. Built and run live in Lodge. Data: nflverse (CC-BY).

We built a digital twin of a Wisconsin paper mill in Lodge, on real power-price, weather and gas data. It forecasts price spikes, plans around them, and cut the bill 28.8% in a 30-day replay (illustrative mill, real data).

Built a world monitor over the weekend with 17 live layers, chokepoint exposure, futures, time slider inspired by github.com/koala73/worldmoni…. Click Hormuz: 60% of transits are tankers, Brent and LNG update live, scrub back three days and watch it move. None of it is bespoke, it's all Lodge primitives. All feeds, geometry, scoring, replay are part of the normal product. Point them at a supply chain instead and you get a different product.

Lineage is one of those things that is not glamorous at all but makes data engineering 10x easier and more trustworthy. Those files on the left are literally screenshots that I uploaded of the spec for what I wanted built. So I can not only can I go back to the individual cell each value was derived from, but even the screenshots the idea for the whole build came from in the first place.

Just made large materializations repairable instead of rebuild-only. Materializations are where Lodge turns raw tables/sources into durable, governed entities (i.e. the thing the rest of the system actually works with). If I find one property out of thousands was mapped incorrectly, I can patch that property instead of rerunning the whole mapping or fiddling with random scripts in standard ETL

I resolved team, coach, league, conference, and player data across 3 source, 15 tables for this cbb ontology. For example, each player is resolved across all tables so that you can see every game they played in, their aggregated stats, all teams they were on, conferences, coaches they played under, etc. Lodge of course keeps granular lineage so that editing the mappings and purging outputs is very easy and every value can always be traced back to the source.




