This article connects two parts of aogl.cn that usually live in separate tabs: the games hub (first-party Steam / console / publisher RSS headlines) and the editorial articles where I archive generative character look-dev. It is a personal Sunday workflow—not a review site, not an art scraping guide, and not affiliated with any storefront.

The Sunday loop in plain language

1. Open games.html after npm run fetch-hub-news refreshes Hot mix.
2. Skim for art-direction-heavy posts (reworks, character trailers, style guides)—ignore patch-note-only drops unless I play the title.
3. Ask a narrow generative question in Midjourney or Gemini: “palette + silhouette” only, not “copy this official key art.”
4. Save explorations under original/<project>/ with dated filenames; reject anything that drifts identity anchors.
5. If a folder matures, publish an honest article (HTML sheet, contact sheet, or video)—examples below.

What I refuse to do

• No store art scraping — thumbnails and key art stay on vendor CDNs; I do not mirror them on aogl.cn.
• No APK / crack indexes — already excluded in games hub editorial rules.
• No “10/10 must play” lists — traffic strategy on this site targets maintainer archives, not generic game rankings.
• No prompt dumps — articles explain decisions; they are not SaaS tutorials.

Case studies already on this site

These folders matured from headline-inspired look-dev into standalone URLs:

• Elena · Poem of the Sea Wind — island priestess HTML sheet + explorations.
• Elena palette & ritual UI — maintenance notes for the same sheet.
• Role-girl card study — white dress contact sheet line.
• Turnaround walk — orthographic + walk-cycle discipline.

None claim to be official fan art for a commercial title; they are practice archives triggered by keeping up with industry visuals.

When RSS is enough without generative tools

If a headline only announces server maintenance, I stop at the hub—no article. The games page is already valuable as a reading rail. Generative work earns a new URL only when I produce inspectable files (PNG, HTML, MP4) and can write a scope paragraph.

Tool directory cross-links

Video look-dev: Runway notes. Coding the sheet: Cursor. Long-form copy edits: Claude. Cited web search when a press release references an unfamiliar engine: Perplexity. Full list: guides index.

Publishing checklist before I add articles/index.html rows

1. Lead paragraph states fictional vs real scope (staging copy, not a listing).
2. English body ≥ 800 words or equivalent depth with figures/embeds.
3. Stable filenames under original/ referenced in figcaptions.
4. Internal links to related demos (window composites, WebGL globes, other character lines).
5. dateModified updated when assets change; changelog entry for humans.

Why this belongs in articles/ not games.html

Hub pages stay short editorial rails with outbound links. Long workflow prose would dilute their purpose and trigger “thin hub” signals. Articles carry the teachable narrative; games.html stays the Sunday headline strip.

What I actually scan for in Hot mix

Not every RSS item deserves a generative session. I treat these as green lights: character redesign blogs, art book announcements, “making of” interviews, stylized trailer drops, or patch notes that include UI mockups. Red lights I skip entirely: server maintenance, anti-cheat-only posts, esports bracket results, and pure sales percentages without visuals. The discipline is boring but it keeps original/ from filling with folders I will never write about.

When a headline is ambiguous—say, “visual update” with no screenshot—I open the outbound link once. If the publisher page is text-only, I stop. I do not guess art direction from a title slug.

Folder naming under original/

Every mature line gets its own directory: original/game-girl/, original/role-girl/, original/character/. Inside, filenames stay chronological: ChatGPT Image YYYY…png for explorations, a single canonical turnaround (ChatGPT-Image2.png or card.png), and optional HTML sheets. I never reuse a folder for a different face—that is how card lines collide on disk and in search snippets.

Before an article exists, the folder must contain at least: (1) one ruler image or sheet, (2) three dated explorations showing a decision, (3) a one-paragraph scope note in my local README (not published until the article ships).

Folders I killed before publishing

Two 2026 experiments never became URLs: a cyberpunk runner silhouette that drifted into copyrighted logo shapes after three frames, and a “fantasy priest” pass that copied a commercial game’s cape cut too closely once I compared side-by-side. Deleting early is cheaper than writing an article that admits mimicry. Those PNGs never entered git.

Contrast with tech.html

The tech hub feeds chip and model headlines. I read it for policy context—export rules, API pricing—but I rarely spawn character folders from it. When tech RSS and games RSS overlap (console SoC news), I file the clip under tech mentally and only cross-link in an article if it changed a tool I use (for example, a local inference speed bump affecting batch exports).

Sunday time budget

Roughly ninety minutes: twenty for RSS skim, forty for one generative pass with narrow prompts, twenty for filing and renaming, ten for deciding “article or not.” If the pass produces nothing keeper-grade, I still commit hub JSON updates but skip generative work—no guilt-driven padding.

Legal hygiene I keep visible

Store key art stays on vendor CDNs. I do not train on scraped key art, do not remove watermarks for publication, and do not label practice folders as official fan works. Articles repeat this boundary so AdSense reviewers see intent, not a warez or rip site.