Building the audit that watches the automations.
A system that checks its own health — and the day it caught its own blind spot.
One person, running on a personal operating system — capture, digest, retrieve. Below: who I am, how the system works, and what it has built. Hand-made, not hand-held.
Telegram in, daily brief out. Every note, article and decision flows through one pipeline — so nothing is lost and everything is findable.
Most people drown in their own inputs — bookmarks they never reread, notes they never find, ideas that evaporate by lunchtime. I treat that as an operations problem, because it is one.
The personal OS has three moving parts. Capture is frictionless: a message to a Telegram bot, a saved bookmark, a voice note. Process turns raw input into clean, linked markdown in a single vault. Digest reads it all back and hands me a styled morning brief.
None of it is magic. It is small, boring automations wired together with judgement about what actually deserves attention. The same instinct I bring to a finance back office: remove the manual step, keep the human one.
A finance professional — audit, fund services and private-equity fund finance — now building the tools, not just specifying them. I build fast with AI, then learn how it works by doing.
Read the full story →Every build starts with research, then a clear scope, then the smallest system that removes the manual step. I would rather do something once and well than thirteen times by hand.
See how I work →A system that checks its own health — and the day it caught its own blind spot.
The gap in audit isn't technology. It's execution — and the easiest wins are in plain sight.
An AI agent, not an RPA bot, cleared a backlog of scanned contracts overnight. The difference was judgment.
A daily heartbeat that confirms every job ran, plus a weekly pass that catches when my own records drift apart.
View →Four-eyes review — the control every finance team trusts — wired into an AI that ships work. Nothing lands without a human.
View →A post-M&A, PE-backed group with no shared system — thirteen-step manual processes turned into one. AI wins on ownership, process and data.
View →