Every page, in one place.
28 published features across five sections, plus everything in Information.
A company that actually rebuilt something, and what AI still could not do.
- Finout turns cloud cost allocation into a reviewable AI workflow
- Qodo brings repository context into monday.com code reviews
- Replicant says CorVel cut injury intake time in half
- Ketryx reports Meta compliance reviews cut from weeks to minutes
- Freed says 20,000 clinicians use its AI workflow for visit notes
- Tamnoon moves cloud alerts through an approved remediation workflow
- OpenOps turns cloud-cost recommendations into workflows
- PaynetEasy gives AI payment data, not payment controls
- Papaya Global reports 3x workers served with flat internal headcount
- How Papaya Global reconciles contractor records before payroll
The operators behind the rollout: what they changed, what they carry, what they refuse to hand over.
Nothing filed here yet.
Two ways of running the same process, side by side, with the trade-offs left in.
The claims that do not survive contact with a real operation, tested against the work.
- Cheap API, expensive agent: where the coding bill grows
- A repeatable workflow does not automatically need an AI agent
- Reddit Isn't a Survey. It Can Still Be Useful Research.
- Can an AI Detector Prove Who Wrote It? No.
- A completed record is not proof of a correct read
- Why Identical Cloud Alerts Can Need Different Fixes
- The Cloud Alert Is Green. Is the Vulnerability Really Gone?
- Read-only AI access is not a security boundary
- The Paycheck Landed. Is Payroll Really Finished?
- Why One Paycheck Can Create Two Payroll Jobs
- Reddit isn't the chess club. It's the best context AI will ever get.
- The roastery that refused to grow up
- A gym for people who never liked gyms
Five ranked pieces of nonsense: fake agents, magic 'skills', demos that die in week two.
Reading
Information
For machines
- sitemap.xml
Every public URL, for search engines.
- feed.xml
RSS: every feature, newest first.
- llms.txt
What this journal is, in plain text.
- llms-full.txt
Full text of every feature, for answer engines.
- robots.txt
What crawlers may and may not read.