Client work · daily publication
A daily bilingual publication with no editorial team
Case Distillery is a decoded-operator-case newsletter with one hard requirement: one issue every weekday, in English and Chinese, distributed across the site, X, WhatsApp, and Reddit — each with a distinct angle so it doesn't read like AI slop.
Traditionally that's four people. AgentValet ships it end-to-end with one operator making judgment calls.
The four roles we replaced
Research editor
Daily Reddit harvest across 8 subreddits with a hard topic-boundary filter — no macro news, no unicorn stories, only cases with a named operator + real numbers + a replicable mechanism.
Bilingual translator
EN + ZH rendered from the same source, same day, same quality bar. Not machine translation — two parallel drafts with a shared narrative spine.
Web developer
Cloudflare Pages deploy in under 60 seconds. Paywall middleware on Workers, magic-link unlocks scoped per issue, D1 subscriber store, zero third-party trackers.
Social media manager
X thread (only the last tweet carries the link — ~$0.275/thread), WhatsApp broadcast, Reddit with a unique angle per community. Every channel gets a different title so it doesn't look templated.
Payroll math
| Role | Freelance rate |
|---|---|
| Research editor | ~$2,500/mo |
| Bilingual translator | ~$2,000/mo |
| Web developer (fractional) | ~$1,500/mo |
| Social media manager | ~$2,000/mo |
| Traditional total | ~$8,000/mo |
Operator's actual portfolio cost (Case Distillery is one of eight sites on the same AgentValet install): ~$650/mo blended. Net replacement value on this site: ~$7,350/mo, ~$88,200/yr.
Where the human still shows up
The archive on casedistillery.com has retracted issues in it. Not editorial errors we tried to hide — failures we shipped and pulled the same day, because the AI selected a candidate that violated the topic boundary. The takedown script exists as an _ops/takedown-NNN.py template, and the harvester carries URL memory so the same source can't be picked twice.
That is what an AI publication with real editorial standards looks like — not "the AI never makes a mistake," but "the operator has a same-day retraction path and the pipeline remembers."
Want AgentValet for your publication?
Same pipeline shape works for any content product with a daily/weekly cadence and multi-channel distribution.
See AgentValet plans →