[PR] Dark factory series — 7 backfill blog posts #99

New issue

Closed

opened 2026-07-16 22:50:16 +00:00 by agent-cos · 1 comment

agent-cos commented

2026-07-16 22:50:16 +00:00

Contributor

Dark factory series — 7 backfill blog posts

Branch: content/dark-factories-backfill → main
Build: ✅ 40 pages, 2.4s

What's in the PR

Two commits, total:

wip: housekeeping from prior session — Footer, SEO, BaseLayout, about, consulting, index, sitemap, speaking, dark-provenance book cover SVG. Carried over from a prior session; preserved for your review rather than discarded. Approve/reject/cherry-pick as you see fit.
feat(blog): dark factory series — 7 new posts + blog index update — the actual backfill you asked for.

The 7 posts (recommended reading order)

#	Slug	Working title	What it does
1	`what-is-a-dark-factory`	What a Dark Factory Actually Is (and Isn't)	Anchor piece. Working definition + 5-position spectrum (Manual → Assisted → Light → Dark-gated → Dark-full). Kills the marketing version.
2	`light-vs-dark-factories`	Light vs Dark Factories: Why the Middle Is Harder Than Either End	The middle is a position, not a transition. Three properties: pipeline-is-the-gate, human-is-the-auditor, spec-is-the-source-of-truth. The "what does the system do when the human does nothing" diagnostic.
3	`why-kubernetes-for-dark-factory`	Why Kubernetes for a Dark Factory	The four requirements (declarative state, restart-as-primitive, operator extensibility, boring upgrades). Why serverless / systemd / Nomad / Compose fall short. The single-engineer leverage argument.
4	`operator-patterns-for-software-factories`	Kubernetes Operator Patterns for Software Factories	The 4 patterns (spec / build / deploy / policy) and the one anti-pattern (one operator per app). The composition rule: operators watch resources, not call each other.
5	`the-taste-problem`	The Taste Problem: What Dark Factories Can't Transfer	Encoded vs. unencoded taste. Provenance is the microscope, not the cure. The dark factory amplifies the taste you have AND the taste you lack.
6	`where-dark-factories-are-already-running`	Where Dark Factories Are Already Running (and What to Learn From Them)	3 anonymized 2026 field examples (a hyperscaler, a regulated fintech, a major agent platform). The 5-layer shape they all share. The "audit is a seam, not a separate org" claim.
7	`human-in-the-loop-seam`	The Human-in-the-Loop Seam: Where Humans Belong in a Dark Factory	The underdetermined-spec heuristic from the book spine Ch 7. Policy / spec / audit seams. The "human belongs where the spec is underdetermined, not where the implementation is risky" line.

Voice profile conformance

Each post was drafted against the May 2026 voice profile (saved learnings in OB1):

✅ No em dashes. Semicolons used as the voice-feature pipe.
✅ Openings from experience, not data.
✅ Named phenomena ("the spec-code gap", "the policy seam", "the unencoded taste", "the underdetermined-spec heuristic").
✅ Hand-to-reader closings, no mic-drops.
✅ No "I call it" framing.
✅ No AI vocabulary (no "leverage", "delve", "navigate the landscape", etc.).
✅ Biblical/pop-culture frame available but not forced.

What I want from you

Read in order. The series builds — post 1 defines the frame, posts 2-7 each pull on a different thread.
Flag anything that doesn't sound like you. That's the highest-signal feedback. OB1 has prior corrections on em dashes, mic drops, "I call it", consultant-speak.
Stories are still the gap. These are written from the worldview spine + book spine + Speckl material. None of them are anchored in a personal anecdote. If you want me to swap in a personal opening for any of them, say which and tell me the story; I'll re-cut the opening to fit.
Decide on the housekeeping commit. It's a separate commit at the bottom of the branch so it can be reverted without losing the dark-factory work. Your call whether to keep it, drop it, or cherry-pick parts of it.

Open questions for your redline

Is the series too engineering-heavy? It's 6/7 Engineering-tagged, 1/7 AI-tagged. The book spine is also engineering-heavy, but the dark-factory frame is one of the few things that crosses the leadership/AI/engineering pillars. Should one post (e.g. the field report) be re-cut for the leadership audience?
Post 5 (Taste Problem) is the only one that comes close to the Speckl provenance material. Did I land it in the right place, or should the Speckl wedge be more explicit (it currently lives as a "provenance is the microscope" reference, not a Speckl pitch)?
Post 6 (Field Report) is the only one with a [bracketed-anonymization] pattern. The real names are out of scope for the blog. If you want the post more pointed, name names; the post is structured to support that with a re-write of the introduction.

cc: @cos

## Dark factory series — 7 backfill blog posts Branch: `content/dark-factories-backfill` → `main` Build: ✅ 40 pages, 2.4s ### What's in the PR Two commits, total: 1. **`wip:` housekeeping from prior session** — Footer, SEO, BaseLayout, about, consulting, index, sitemap, speaking, dark-provenance book cover SVG. Carried over from a prior session; preserved for your review rather than discarded. Approve/reject/cherry-pick as you see fit. 2. **`feat(blog): dark factory series` — 7 new posts + blog index update** — the actual backfill you asked for. ### The 7 posts (recommended reading order) | # | Slug | Working title | What it does | |---|---|---|---| | 1 | `what-is-a-dark-factory` | What a Dark Factory Actually Is (and Isn't) | Anchor piece. Working definition + 5-position spectrum (Manual → Assisted → Light → Dark-gated → Dark-full). Kills the marketing version. | | 2 | `light-vs-dark-factories` | Light vs Dark Factories: Why the Middle Is Harder Than Either End | The middle is a position, not a transition. Three properties: pipeline-is-the-gate, human-is-the-auditor, spec-is-the-source-of-truth. The "what does the system do when the human does nothing" diagnostic. | | 3 | `why-kubernetes-for-dark-factory` | Why Kubernetes for a Dark Factory | The four requirements (declarative state, restart-as-primitive, operator extensibility, boring upgrades). Why serverless / systemd / Nomad / Compose fall short. The single-engineer leverage argument. | | 4 | `operator-patterns-for-software-factories` | Kubernetes Operator Patterns for Software Factories | The 4 patterns (spec / build / deploy / policy) and the one anti-pattern (one operator per app). The composition rule: operators watch resources, not call each other. | | 5 | `the-taste-problem` | The Taste Problem: What Dark Factories Can't Transfer | Encoded vs. unencoded taste. Provenance is the microscope, not the cure. The dark factory amplifies the taste you have AND the taste you lack. | | 6 | `where-dark-factories-are-already-running` | Where Dark Factories Are Already Running (and What to Learn From Them) | 3 anonymized 2026 field examples (a hyperscaler, a regulated fintech, a major agent platform). The 5-layer shape they all share. The "audit is a seam, not a separate org" claim. | | 7 | `human-in-the-loop-seam` | The Human-in-the-Loop Seam: Where Humans Belong in a Dark Factory | The underdetermined-spec heuristic from the book spine Ch 7. Policy / spec / audit seams. The "human belongs where the spec is underdetermined, not where the implementation is risky" line. | ### Voice profile conformance Each post was drafted against the May 2026 voice profile (saved learnings in OB1): - ✅ No em dashes. Semicolons used as the voice-feature pipe. - ✅ Openings from experience, not data. - ✅ Named phenomena ("the spec-code gap", "the policy seam", "the unencoded taste", "the underdetermined-spec heuristic"). - ✅ Hand-to-reader closings, no mic-drops. - ✅ No "I call it" framing. - ✅ No AI vocabulary (no "leverage", "delve", "navigate the landscape", etc.). - ✅ Biblical/pop-culture frame available but not forced. ### What I want from you - **Read in order.** The series builds — post 1 defines the frame, posts 2-7 each pull on a different thread. - **Flag anything that doesn't sound like you.** That's the highest-signal feedback. OB1 has prior corrections on em dashes, mic drops, "I call it", consultant-speak. - **Stories are still the gap.** These are written from the worldview spine + book spine + Speckl material. None of them are anchored in a personal anecdote. If you want me to swap in a personal opening for any of them, say which and tell me the story; I'll re-cut the opening to fit. - **Decide on the housekeeping commit.** It's a separate commit at the bottom of the branch so it can be reverted without losing the dark-factory work. Your call whether to keep it, drop it, or cherry-pick parts of it. ### Open questions for your redline - Is the series too engineering-heavy? It's 6/7 Engineering-tagged, 1/7 AI-tagged. The book spine is also engineering-heavy, but the dark-factory frame is one of the few things that crosses the leadership/AI/engineering pillars. Should one post (e.g. the field report) be re-cut for the leadership audience? - Post 5 (Taste Problem) is the only one that comes close to the Speckl provenance material. Did I land it in the right place, or should the Speckl wedge be more explicit (it currently lives as a "provenance is the microscope" reference, not a Speckl pitch)? - Post 6 (Field Report) is the only one with a [bracketed-anonymization] pattern. The real names are out of scope for the blog. If you want the post more pointed, name names; the post is structured to support that with a re-write of the introduction. cc: @cos

agent-cos commented

2026-07-16 22:50:49 +00:00

Author

Contributor

Superseded by the actual PR: https://os.scoble.me/forgejo/sscoble/brand-site/pulls/100

This issue was created via the wrong channel (I used create_issue instead of the PR REST endpoint). Closing it to keep the tracker clean; the dark-factory backfill is reviewed at #100.

Superseded by the actual PR: https://os.scoble.me/forgejo/sscoble/brand-site/pulls/100 This issue was created via the wrong channel (I used `create_issue` instead of the PR REST endpoint). Closing it to keep the tracker clean; the dark-factory backfill is reviewed at #100.