Content Systems · Rob + Lam, jointly
Content infrastructure that publishes itself
Content strategy fails at the production step: the plan is right and the publishing never keeps up. We build the machine — research agents, editorial rules, drafting pipelines, human review gates, and distribution — and then we run it, so publishing happens on schedule whether or not anyone had a good week. The build is the first month. Watching it and improving it is the relationship.
This is the joint service: Rob Garner sets the editorial and search strategy — what machines and readers each need to see — and Lam Hoang builds and operates the pipeline that produces it. Everything below runs on our own properties first. The work page has the numbers; the podcast is the machine talking about itself.
What this work looks like
Automated newsrooms
Source monitoring, newsworthiness scoring, drafting under editorial rules, and multi-channel syndication. Descout, our own newsroom, has published this way for 313 consecutive days across 160+ entities. The pipeline watches sources continuously, scores what's actually worth covering, drafts to a stylebook enforced in code, and syndicates — with a person able to step in at every stage. The editorial rules are the real product; the automation just refuses to forget them at 11pm on a Friday.
Podcast pipelines
Ideation to episode to six-platform distribution, automated. The WorkHacker Podcast ships on exactly this system — 36 episodes across Podbean, Spotify, Apple, Amazon Music, iHeart, and Castbox, and the proof is in your podcast app. Topics come out of a research loop, production runs on a fixed cadence, and distribution is a pipeline rather than a Friday-afternoon chore. The show is also where the thinking behind these systems gets argued in public, episode by episode.
Closed-loop content engines
Research agents feed a living knowledge base; the knowledge base feeds generation; performance feeds back into research. Content that compounds instead of piling up. The engine behind Oase's marketing runs this loop through a twelve-stage pipeline from research to publish. The closed loop is the point: most content operations are open-ended — publish and hope. Here, what worked reshapes what gets researched next, and the knowledge base becomes an asset that outlives any single article.
Internal linking & site optimization
A private engine that proposes internal links with natural anchor text, routes every change through human approval, and measures its own lift in Search Console — before/after, per link. Internal linking is architecture, and architecture is meaning: how pages connect tells retrieval systems what a site is an authority on. The strategy behind that sentence is the core of our agentic SEO practice; this is the tool-assisted version, applied safely — snapshots and undo included.
Editorial oversight, ongoing
Sources move, rules go stale, models change. We read what the system publishes the way an editor does, adjust the rules behind it, and tell you what shipped and what it earned. Nothing runs unwatched. It's the same managed-automation discipline as our engineering engagements: the machine does the volume, people do the judgment, and the monthly report says what happened in plain numbers.
How an engagement runs
Editorial rules first
Before any pipeline exists, we write the stylebook the machine will be held to: voice, sourcing standards, what counts as newsworthy, what's off-limits. Rob's editorial strategy plus your domain knowledge, turned into rules code can enforce — because a pipeline is only as good as what it's told to refuse.
Build and calibrate
The pipeline goes up — sources, scoring, drafting, review queue, distribution — and drafts against the rules while a human reviews everything. Wherever a draft diverges from what your editor would approve, the rules change until it doesn't. Nothing publishes without a sign-off in month one.
Publish on schedule
The system publishes on cadence whether or not anyone had a good week — that is the point of building it. Human review stays at the gates that matter, and the share of drafts that pass review untouched becomes a metric we report and push upward.
What shipped and what it earned
Pieces published, review pass-rates, Search Console movement, citations and syndication pickup where measurable, and what we changed in the rules. Content systems drift; the monthly report is where drift gets caught before readers catch it.
Proof, with numbers
36
episodes · six platforms
The WorkHacker Podcast
Our own show, produced by the pipeline described above — ideation, production, and distribution across Podbean, Spotify, Apple, Amazon Music, iHeart, and Castbox, on a cadence no ad-hoc process was keeping.
313
consecutive days · 160+ entities
Descout
An automated newsroom in production: monitoring, newsworthiness scoring, drafting under enforced editorial rules, and syndication — the longest-running proof we have that a machine can hold a publishing schedule.
Runs on our own properties first · human-approved, reversible
A private internal-linking engine
The site-optimization side of content infrastructure: proposes links with natural anchors, routes everything through human review, applies changes with snapshots and undo, and measures its own Search Console lift.
Why us
- We run these systems on our own properties first, and we have never stopped tuning them. The podcast, the newsroom, the content engine behind Oase's marketing — all live, all automated, all inspectable.
- Strategy and engineering in the same room: the search strategist defines what machines need to see, the engineer builds the pipeline that produces it. Both of them are on the about page.
- Human-in-the-loop by design: the machine drafts unattended, and a person signs off before anything ships.
- Measurement is part of the build, not an afterthought: every system reports what it produced and what that produced in return.
The WorkHacker Podcast · Descout · Oase content engine · all in production
Common questions
07 Questions
Can AI-generated content actually rank and get cited?
Structured, sourced, editorially governed AI content can rank and get cited; unedited bulk output mostly cannot. The difference is the system around the model: sourcing standards, editorial rules enforced in the pipeline, human review gates, and site architecture that makes each page unambiguous. Our own newsroom has published for 313 consecutive days under those constraints — the constraints are why it works.
What is an automated newsroom?
An automated newsroom is a pipeline that monitors sources, scores what's newsworthy, drafts coverage under enforced editorial rules, and syndicates it — with humans at the gates that matter. It isn't a scheduling tool or an AI writer; it's the whole editorial operation, encoded. Descout, ours, covers 160+ entities this way.
Does a human review everything before it publishes?
In month one, yes — every piece passes a human gate while the rules calibrate. After that, review concentrates where judgment matters: sensitive topics, new formats, anything the scoring flags as unusual. The system earns autonomy piece by piece, and the share of drafts passing review untouched is a metric we report. Nothing ever runs unwatched.
How is this different from hiring a content agency?
An agency sells hours; we build and operate a system, so output doesn't scale with headcount. The pipeline does the monitoring, drafting, and distribution; editorial judgment — ours and yours — goes only where it's needed. It publishes on schedule through vacations and bad quarters, and everything it does is inspectable: rules, drafts, decisions, numbers.
How do you measure whether the content works?
Every system reports what it produced and what that produced in return. Concretely: pieces shipped, review pass-rates, Search Console impressions and queries, rankings and AI-answer citations where they apply, and syndication pickup. Baselines are taken before launch so lift is attributable to the system. Measurement is part of the build, not a slide assembled at renewal time.
What does the machine do on a normal day?
It watches sources, scores candidates against the newsworthiness rules, drafts what clears the bar, queues anything gated for human review, publishes what's approved, and syndicates — then logs all of it. On a good day nobody touches it. On an interesting day, the review queue is where the interesting part waits.
What happens when output quality drifts?
Drift is expected and budgeted for — sources move, models change, rules go stale. Because we read what the system publishes the way an editor does, drift shows up in review notes and pass-rates before it shows up in public. The fix lands in the rules layer, versioned like code. Catching this early is what the ongoing engagement is for.
Working together