Personal-Branding Content Engine — Build Spec

The stream: Personal-Branding is one of the crucial arms of Arijit's AI-OS. Goal: build his public brand by publishing consistently — ≥2 posts/week for a year (~100+ pieces) — without it becoming a time sink. The blog is the mother asset; LinkedIn/X/Slack/WhatsApp pieces are lead magnets that drive traffic back and are measured. Target framing (from strategy): the deliverable is the machine, followers are the lagging indicator. Reframe of the 10k/60d goal: build the engine, measure at day 30/60.

This job: build the end-to-end engine + the skills + the dashboard + the blog system + the backend. Reuse-first. Everything inherits DESIGN-SYSTEM.md. Commit to git + vault under AI-OS / Personal-Branding.

1. The end-to-end pipeline

Research/topic-hunt → Score & pick → Brainstorm/outline → Draft → Six-pass edit (Arijit's voice) → [GATE: Arijit approves] → Diagrams/images → Publish (blog system) → Atomize (carousel + LinkedIn/X/Slack/WhatsApp) → Schedule/remind → [Arijit posts] → Measure (Umami + LinkedIn native) → Govern (winning topics/formats feed back).

1b. Topic-hunting engine (DECIDED 2026-07-07 — the sources + the in-loop flow)

Arijit stays IN THE LOOP at topic selection (his explicit ask). Weekly scan across ~10 sources → ranked shortlist → he picks.

Sources the engine watches: - ⭐ Arijit's own Claude/Codex transcripts — mine what he's actually building + his working style/opinions → authentic, evidenced, un-copyable topics (the Trust post came exactly this way, from the Phase-0 evidence-ledger). This is the differentiator. Build a transcript-miner. - Frontier labs: Anthropic, OpenAI, Google/Gemini blogs + release notes + announcements (RSS + scrape) → "what execs should understand about X." - GitHub trending repos + notable AI/agent releases (API). - X, YouTube, Hacker News, Reddit — reuse last30days (already multi-source over these). - Arijit's vault + projects (PRISM/Scout/CoE) → "what the data said" posts. - Comment-mining — every question in his post comments = a future topic. - (optional/later) arXiv, curated newsletters.

Flow: topic-hunter (weekly, scheduled via Hermes/cron/loop) runs the scan → each candidate = {topic, why-now trigger, winning angle via his thesis lens, evidence, source, pillar} → topic-scorer ranks vs thesis + pillars + "nobody-made-it-digestible-for-execs" angle → surfaces top 5–8 shortlist → Arijit thumbs-up/down/notes (the loop) → pick feeds the writer; his choices train the scorer. Govern: traction (Umami + LinkedIn) re-weights sources/angles.

1c. Voice engine (DECIDED — use the existing skill)

Voice = the tone-of-voice skill (v2: built from 26 real sent emails + 2 verified Medium essays; Register 3 = long-form is the one for blogs/LinkedIn — narrative cold-open, rhetorical-question engine, coined frameworks, evidence-dense, philosophy+pop-culture range) + the WritingSOPs long-form methodology. The blog-writer loads both. The skill self-deepens: every real Arijit-authored piece read → extract + update it. No new paragraphs required (more only sharpens it).

2. Skills to build (reuse-first — wrap what exists)

Honesty gate (non-negotiable, learned): every asset fact-checked vs the real evidence before it ships. Derivatives inherit the blog's "fabricate nothing" thesis. (This session caught Fable inventing "27 tics" + a fake force-push in the carousel/article.)

3. Blog system (finish the buildout)

Currently 1 hand-rolled page (/1). Need a structured multi-page site: index / chapters / per-post pages / nav / tags / RSS / back-and-forth navigation. [REC default: Astro static site] — Markdown-authored posts, design-system as components, builds to static served off chowmes/Caddy, scales to 100 posts, unifies with the planned Astro brand hub. Migrate the current Trust post in as post #1.

4. Dashboard

The Content layer of the AI-OS cockpit (ties into aios-cockpit-build): a pipeline board — Topic queue → Drafts → Approved → Scheduled → Live — with metrics (Umami visits, LinkedIn impressions, follower trend, which channel drives traffic). This is Arijit's content command center.

5. Backend / infra

• Content store: the Obsidian vault (SSOT) — posts, topic queue, voice profile, metrics log.
• Publish: blog build → chowmes/Caddy (scratchpad.chowmes.com or the future brand domain).
• Analytics: Umami (installed this session, boxed on chowmes, website created, tracking-id 7c67209b-8bd6-48f5-b68e-f7fca9c9be34) — needs a public route + the tracking tag (last wiring step). LinkedIn impressions = native.
• Posting: LinkedIn API can't auto-post carousels to a personal profile → content-factory + Arijit posts (prep everything, he taps post). Paid scheduler (Taplio/Buffer) optional later.

6. Design system

Everything inherits DESIGN-SYSTEM.md (Fresh Editorial: warm off-white, Fraunces/Newsreader, fresh pastels, pinned-photos, color-morph bg, palette picker). Carousel re-skins from Evidence-Noir dark → Fresh Editorial.

7. Learn-from-others (don't reinvent — Arijit's directive)

Before building carousel-creator + the posting cadence, research how others already solved this: LinkedIn carousel tools (Taplio, AuthoredUp, Contra, Canva carousels), top "build-in-public" creators' systems, YouTubers who teach carousel/LinkedIn growth. Extract the proven patterns (hook slide, dwell-time format, "link in comments," document-post specs, posting times) → adopt, don't invent.

8. DECISIONS (all locked 2026-07-07)

1. Blog stack: ✅ Astro static site (Markdown posts → static HTML; nav/tags/RSS auto; design-system as components; served off chowmes/Caddy; scales to 100 posts; unifies w/ the Astro brand hub). Migrate the Trust post in as post #1.
2. Autonomy: ✅ auto-to-draft, Arijit gates twice (topic pick + final approve). PLUS he stays in the loop at topic selection — engine surfaces a shortlist, he picks (see §1b).
3. Tonight: ✅ ship carousel + 1 LinkedIn post → the live blog. Carousel RE-SKINNED to the Fresh Editorial design system (off Evidence-Noir); design its content (slides, boxes, graphics, data, hook, CTA); LinkedIn post written in tone-of-voice Register 3.
4. Voice: ✅ the tone-of-voice skill is the engine (see §1c) — self-deepening, no new input required.

9. Pillars (mother-blog roadmap, from strategy doc)

1. ✅ Trust architecture (shipped). 2. Loop/harness engineering (next — "make it digestible for execs"). 3. The retriever thesis ("model isn't the bottleneck"). 4. Enterprise AI operating model (PRISM/CoE). Unifying thesis: "The bottleneck in AI is not the model — it's the system around it."

10. Sequence (fresh context)

P0 lock design-system + finish Umami wiring · P1 blog system (Astro) + migrate Trust · P2 carousel-creator (re-skinned) + tonight's LinkedIn post · P3 topic-hunter + scorer + writer (the auto-pipeline) · P4 atomizer + post-packager · P5 dashboard + govern loop. Each phase shippable.

Related

• DESIGN-SYSTEM
• index