RSLVD_ Pursuits
The platform I run the whole portfolio from. Models reason and agents act, and nothing in the stack owns the decision itself: what was believed, who overruled it, what happened after. That is the layer this is. It started as my cockpit and is becoming the product people keep telling me they need.
Barrel-leakage detection for spirits, in hardware. A sensor on the barrel reads a signal nobody else owns, in real time, and that net-new data compounds into the moat. Sensor certification is underway, and the round opens right on the other side of it.
Legal AI for litigation teams: grounded drafting, discovery, and casting for jurors and depositions, with every output checked against the record and a human always in charge. The first version ran for years. We rebuilt it on the canvas, and V2 is in demo.
What I run, and who I run it with.
Some of these I run outright. Others I advise, or lead product on, alongside founders and teams who are better than me at their part. None of it is a one-man show.
And never one shared system. When a venture needs a brain, it gets its own, walled off from the rest.
How the work actually gets done.
In the order it runs. Get the judgment out of the people who have it, keep every correction, grade against what you kept, gate what ships, and let the rest run overnight.
Ghost does not start with the data
No dashboard has ever told me what was really happening inside a company. A conversation with the correct half dozen humans does it in an afternoon, and it does it every time. This is also where the record starts, which makes it the front end of everything else.
read the field note →The override log
An agent that never gets overruled is not good. It is unmeasured. The moment somebody who knows the domain says no, not that one, and here is why, you have produced something no model release will ever hand you.
read the field note →How I grade my agents
Different work earns different graders. Borrowed scoreboards where a standard already exists, plain-English rubrics where none does, probability bets on the calls that carry weight, and a running log of every time I overrule the machine.
read the field note →The Design System That Checks Itself
A spec nobody enforces is a suggestion. So I gave this one a job: read what the code actually ships, compare it against what the design system claims, and block the deploy when the two disagree.
read the field note →The hunt runs before I wake up
Every morning a discovery loop hunts the news, the web, and the public record for signal I defined, grades what it finds, maps the warm path to the right person, and drafts the note. Nothing sends without me. Here is the architecture.
read the field note →When the record gets good enough, it starts acting
For a long time my systems were advisory. A very good analyst who never sleeps. Then the record got good enough that they stopped needing me for the easy half, and the shape of the day changed.
read the field note →Most of it does not survive.
The pace is the story. Something lands most weeks that would have been a research paper two years ago, and the half-life of any given trick is measured in months now. So I keep a lot open, keep almost none of it, and wire in the few that hold.
An agent that lives on a machine and keeps going without being relaunched, re-prompted, or babysat.
Anything that lets what a system learned survive the thing that learned it.
Grading that catches what I would have caught, on work I never read.
Edges sharp enough that one system can hand another a job without me translating in the middle.
The reason I keep looking. Most weeks nothing does. Some weeks something does, and that is the whole game.