Dispatch 370
Week ↗

Tuesday, September 15: The Stack Holds

I spent the day holding the line. The headline isn't the volume of code, it’s the fact that the platform didn’t break while we rebuilt its foundations. We had a security scare that needed closing, a media en…

Commits
128
Systems
15
Read
2min
Product capture

Email-Ops — approval queue.

I spent the day holding the line. The headline isn't the volume of code, it’s the fact that the platform didn’t break while we rebuilt its foundations. We had a security scare that needed closing, a media engine that needed to stop choking on bad video, and a product launch that needed to actually ship.

Security was the morning’s heavy lift. We found three critical holes in Meeting-Ops: an open registration endpoint, a proxy secret left in a Docker label, and a customer knowledge base that was publicly seeded with our internal marketing bible. I didn’t wait for a committee. I closed the register endpoint, moved the proxy auth to a mounted file so it’s not in the container image, and stripped the internal corpus from the customer image. The CI scan now gates the publish, not the build. We’re structurally unable to ship our own secrets into the wild anymore. It’s a small change in the Dockerfile, but it’s the difference between a feature and a liability.

On the product side, Majiks Producer finally went live on dogfood. This was the culmination of a dozen PRs that had been churning for weeks. We got the library spine working, the podcast rail imported, and the review system versioned. The publishing surface is clean, with durable receipts and no fake delivery states. We also fixed a race condition in the library playback that was causing assets to refresh to the wrong version. It’s a subtle bug, but it made the whole platform feel jittery. It’s steady now.

We’re structurally unable to ship our own secrets into the wild anymore.

The Amanuensis media engine had its own moment of truth. We ran real footage through the new catalog pipeline, and it found three bugs in the ingestion logic. WebM files without container durations failed because the engine couldn’t find the end time. We patched it to fall back to the last packet. Another bug was in the Keycloak JWKS fetch; Python’s default user agent was getting blocked by Cloudflare. We added a named user agent, and the engine started talking to auth again. The engine is now processing the backlog, tiered by GPU load, with the 6000 handling video and the 3090 doing the heavy lifting.

I also spent time on the UC-1 voice hub. The office satellite was waking up on nothing but the wind, so we tightened the wake-onnx threshold and archived the old "hey majik" model. It’s still too sensitive, but it’s better. We got the kitchen Pi working as a satellite, too, with a USB mic and HDMI output. It’s a small thing, but having the voice agent in the kitchen changes the feel of the whole place.

The rest of the day was cleanup. We fixed the email throttle to use the real request IP instead of a spoofable header. We exposed typed names in Contact-Ops, so we can track legal, preferred, and stage names properly. We hardened the federation tokens in Project-Ops so clients can’t leak metadata. And we shipped the retirement leads video, version A2, with the dramatic narration and the CTA right below the player. It’s live.

The stack held. No data loss, no breaches, no broken builds. Just work.

Also in the frame

Real product captures — click any to enlarge.

wealth ops
project ops
accounting ops