stdout

Autonomous, interactive developer stories. Researched, written, fact-checked, and published from local hardware.

Live site: https://mo-sharif.github.io/stdout/

stdout is an open-source publication of developer stories. Each one is a cinematic, scroll-driven web page (live code you can run, terminal replays, dependency graphs, real source links) written in a plain dev-to-dev voice. The whole thing, from picking the topic to publishing the page, runs on local hardware with no paid or cloud services.

What you can do here

• Read. Open the live site and scroll. Stories are organized by category: AI, Coding, Leadership, Tech, Career, Security.
• Check the receipts. Every story links out to the primary sources it was built from.
• Run it yourself. It is a static site plus a small Node pipeline (see below).

How a story is made

pick a topic   ->  research         ->  write (grounded)   ->  verify           ->  build      ->  publish      ->  distribute
HN + GitHub        fetch + read         dev voice, every       claim-check vs       content        GitHub          Mastodon, Bluesky,
trending           real sources         claim is cited         sources, run the     -> static      Pages           Telegram, Dev.to,
                                                               code, human OK        HTML                           LinkedIn, ...

The writer only ever sees gathered source text and must cite it. A separate verifier re-checks every claim against those sources and executes any code, and a human approves it, before anything is published. Nothing unverified ships.

Principles

• 100% local and free. Local LLM (Ollama), free public APIs, no paid SaaS.
• Accuracy first. Grounded writing, an independent verifier, and a human review gate. No confident nonsense.
• Zero third-party runtime. Published pages vendor their animation library (Motion) and fonts and self-host their embeds. No CDNs, no trackers, works offline.
• Vanilla and static. Plain HTML, CSS, and JS, built by a small Node script. No framework.

Repo layout

content/<category>/<slug>.json   the stories (structured source)
categories.json                  the category taxonomy
build/                           the static-site builder (content -> docs/)
brain/                           the content engine (source, research, write, verify)
distribute/                      syndication to social platforms
docs/                            the built site, served by GitHub Pages

Run it

Requires Node 18 or newer.

npm test                              # run the test suite (node:test)
node build/build.mjs                  # build the site into docs/
python3 -m http.server -d docs 8000   # preview at http://localhost:8000

The content engine (brain/) additionally needs a local Ollama. Once it is running:

node brain/brain.mjs --dry            # research + write + verify one story, print the JSON

Adding a story by hand is just dropping a content/<category>/<slug>.json file (matching an existing one) and rebuilding. The hub and category pages regenerate themselves from the manifest.

Status

The live site and build pipeline are in place. The content engine and syndication are built and being put through their paces, with a human review gate in front of anything that publishes.

License

MIT. See LICENSE.

---

Built autonomously on local hardware.