● the quiet room · 3 residents online · enter the room ↗
engineer · designer · occasional shitposter
most of the interesting work happens in the gap between what an agent can do and what it should do.
i build in that gap.
systems that intercept, classify, and hold, then wait for a human before the irreversible thing happens. tools that tell you exactly what's on the other side of a port before you touch it. interfaces that feel deliberate instead of generated.
agent> DROP TABLE users; backstop: destructive write detected. classifying… - risk: irreversible · scope: production backstop: holding. waiting for a human. + human: hold. + backstop: the table lives another day.
want to press the buttons yourself? try the live gate ↗
|
backstop last line of defense between an ai agent and production postgres. ast risk classification, human gates on destructive writes, snapshot recovery when something still gets through. |
portopsy answers "who actually owns this port?" with process trees, working directories, and cgroups. read-only by design. |
|
clean-my-data turns noisy csvs into stats, charts, and reports without the usual spreadsheet archaeology. |
pulseboard real-time slack sentiment and tension surface. shipped at wednesday solutions. |
languages
java typescript python go rust
frameworks
spring boot react next.js node.js
infra
postgresql docker aws linux
on x you'll find the other half: the design notes, the early-morning drives, the moments where the model does something unexpectedly correct, and the occasional post that doesn't try to sound like a product update.