nikom.workIndependent engineering studio

Go / Rust / PublishingWorking private system · acquisition controls implemented

A publishing control plane for reviewed releases and controlled acquisition

A private control plane turns keyword exports and source evidence into reviewed multilingual static releases, then records site-level acquisition experiments and outcomes.

8 renderer architectures · validated 50,000-page stream · 24 versioned techniques; only 2 transports automated

The job

What needed to change

At scale, generating a page is the easy part. The operation fails when evidence, review state, media, cost, routes, deployments, experiments, and corrections cannot be inspected together after publication.

The hard part

The constraint shaping the build

Keep evidence and human approval in the publishing path, keep public releases deterministic, and measure post-release work without describing a catalogue of possible tactics as an autonomous growth engine.

System flow

How the pieces connect

  1. Keyword and source library
  2. Bounded production packs
  3. Evidence and media review
  4. Scheduled multilingual release
  5. Go manifest to Rust renderer
  6. Observe and run controlled acquisition
Implemented views

Rendered examples from the build

Neutral publishing fixture rendered in the Radiant architecture
Radiant: one of eight implemented public architectures, shown with neutral source material.
Neutral publishing fixture rendered in the Signal architecture
Signal: the same accepted source library in a different public experience.
Neutral publishing fixture rendered in the Atlas architecture
Atlas: the same facts as a dense, crawlable reference experience.
Matching system

The matching system explains who it is for, what a first release includes, and the rough cost. Supporting links show the implementation evidence.

Programmatic publishing systemSee the configurable system for source intake, review, scheduled releases, static rendering, and post-release controls. Acquisition controls after releaseThe technique catalogue, experiment records, and honest automation boundary.
Build decisions

How I built it

  1. Keep source evidence, generation inputs, outputs, costs, media, and approve, hold, or reject state as separate records.
  2. Build bounded production packs rather than asking one model call to reason over an unbounded site.
  3. Generate a Go release manifest, stream it through a Rust renderer, and activate the validated output atomically.
  4. Record site opt-ins, budgets, pauses, estimates, experiments, deliveries, exposures, outcomes, and URL observations separately.
  5. Keep publishing and acquisition approval explicit; a completed model job or listed technique never acts by itself.
Verification

What I verified

  • 50,000-page streaming renderer validation
  • 50,000-row cannibalization-audit persistence diagnostic: earlier one-run 6,558 ms; newer three-run median 1,010.058 ms
  • Eight public renderer architectures; three neutral previews retained here
  • Twenty-four versioned acquisition techniques with immutable estimate and experiment records
  • Automated IndexNow and XML-RPC delivery through a dedicated Rust distribution worker

Current state: The control plane and large renderer are tested. IndexNow and XML-RPC are the only automated acquisition transports; the other 22 techniques remain manual or proposal-only. The recorded 50,000-row persistence diagnostic is not a renderer or end-to-end release timing. No traffic, ranking, lead, or revenue outcome is claimed.