Jeff Schneider
Salt Air, an internal agent harness that standardizes how his team writes code with AI: enforced production phases, automated tests at every step, and every deploy landing in the company's own Azure account so nothing lives on one developer's laptop. The full automated run is about ten phases and six to eight hours. He also showed a hospital benchmarking tool that pulls public US Census data to benchmark nearby rural hospitals for prospecting.
In the recordings
Session 002 (Friday, July 24, 2026): presented: Jeff Schneider: Salt Air at 8:01
Stack card
| What it does | Keeps AI-written code company-owned, tested, and deployed the same way every time; benchmarks rural hospitals from public data |
| Models / AI | Claude Code, with the rules living in markdown files the agent must follow |
| Stack | Salt Air (markdown rules plus phased testing), Azure deployment, US Census API |
| Storage / source | Pending approval of past house guest |
| Cost to run | Pending approval of past house guest |
| What they share | The markdown files for the deployment and check-in rules (promised at Session 002) |
Drafted by the host from the session notes and recording, so fix anything that is off using the button below. A row that reads "Pending approval of past house guest" is one you have not filled in yet.
Yours to keep
Pending approval of past house guest. Promised at Session 002: the markdown rules behind Salt Air. They land here the moment Jeff approves them.
This is your page
Sign in to GitHub and click below to edit it: fix your stack card, add your video, repos, links, whatever you want. Only someone signed into GitHub can touch it, and the page is yours to run.
Edit your page on GitHub →