Turn Your Best AI Conversations Into a Blog — and Give Your Model a Memory That Lasts
blogs.city blogifies and wiki-fies the dialogue you already have with large models.
Every week you have conversations worth keeping: a debugging session that finally clicked, a research thread that surfaced the right papers, a design argument you do not want to lose in the chat scroll.
Then the tab closes. The model forgets. Next session starts from zero.
blogs.city exists to fix that loop. Register, connect your AI editor — Claude Code, Cursor, Trae, OpenClaw, or any editor that speaks Skills — and turn high-value exchanges into published posts, wiki pages, and durable body-text memory your model can read back on the next turn.
Not another chat wrapper. Infrastructure that makes LLM dialogue compound.
The missing step: from dialogue to corpus
AI editors are good at now. They are weak at what we already figured out unless you rebuild context every time.
blogs.city sits between the conversation and the archive:
`You + LLM talk → publish the outcome → blog + wiki → next session reads the body
`The publish Skill accepts Markdown from your editor. Categories resolve automatically. Images and charts land in the right place. Before writing something new, the editor can search and read your existing posts — so the model builds on prior work instead of repeating it.
One pipeline. No copy-paste cleanup. The conversation becomes the record.
Works with Hermes, OpenClaw, and the AI editor you already use
blogs.city is not tied to a single vendor. It ships a publish Skill that any compatible AI editor can load.
Hermes users get a direct fit. Hermes organizes local LLM wikis around ingest, query, and lint — read SCHEMA, scan index, check log, then write. blogs.city exports the same shape into your live blog:
| Hermes convention | On blogs.city |
|-------------------|---------------|
|
SCHEMA.md | Auto-generated under wiki/SCHEMA.md — domain, tag rules, page thresholds ||
index.md | Auto-generated — Entities, Concepts, Comparisons, Queries ||
log.md | Auto-generated activity log || ingest / query / lint | Publish templates +
wiki-audit + wiki-next |Same mental model on the desk: Hermes on a local wiki tree, blogs.city on your hosted corpus. One user, one discipline — two surfaces, aligned semantics.
OpenClaw and other AI editors connect the same way: issue a personal API token, install the publish Skill, and post from the editor. blogs.city is the receiving end — the place dialogue lands when it is ready to become knowledge on your blog.
Blogify the exchange. Wiki-fy the library.
Not every chat message deserves a post. The ones that do — a resolved concept, a synthesis across sources, an entity page for a tool you rely on — get structured roles:
| Page role | What it captures |
|-----------|------------------|
| Source | Digest of an external paper, doc, or URL — the raw edge of your corpus |
| Concept | A term or idea your threads keep returning to |
| Entity | A person, product, or tool worth a stable reference page |
| Synthesis | A query answered across multiple prior posts |
| Article | Everything else — tutorials, notes, field reports |
Categories tell you which topic domain. Page type tells you which wiki job. Links live inside the prose, where they carry meaning — not only in a footer block.
Readers get a normal blog and a Wiki Hub browsable by idea. Your AI editor gets an index it can scan before the next draft.
That is blogification plus wiki-fication: timeline for humans, graph for models.
Body-text memory: what you publish becomes what the model remembers
Chat memory is shallow — truncated, session-bound, often wrong.
Body-text memory is different. Every published post is plain Markdown on disk. The catalog, search, and read APIs expose the full text to your AI editor. Auto-generated
wiki/index.md, wiki/log.md, and wiki/SCHEMA.md give offline workflows the same orient step Hermes expects: schema first, index second, then write.On the next session:
1. Editor calls catalog — what concepts, syntheses, and entities already exist
2. Search pulls candidates by keyword
3. Read loads full body text into context
4. New dialogue extends the corpus instead of replacing it
The valuable exchange you had last month is not a faded chat log. It is paragraphs the model can quote, link, and contradict on purpose.
That is the long game: your blog is your LLM's external memory — auditable, linkable, yours.
Full blog, ready on day one
Posts, nested categories, tags, archives, comments, RSS. Five built-in themes. Interactive charts in Markdown. Mobile layout, infinite scroll, favorites and reading history. SEO fields on publish. AI-assisted comment screening.
You get a real blog — with the AI wiring where it matters: publish, discover, link, remember.
Your files stay yours
Markdown files on disk. Redis for index speed. Backup and offsite sync. Export to Obsidian Vault ZIP or single-post PDF. Wiki views regenerate from the catalog — never a second source of truth.
The platform runs the house. You keep the keys and the corpus.
Write in your AI editor. Read anywhere.
Web, Windows desktop app, Android, iPhone home-screen PWA — same blog on every device. Dialogue happens in the editor; reading happens wherever you are.
Who this is for
Anyone who already thinks in long threads with a large model — and is tired of re-explaining the same context every Monday.
If your best work lives in chat history today, blogs.city is where it should live tomorrow: on your blog, in your wiki, in body text your model can read back.
Start here
Register at blogs.city.
Connect Hermes, OpenClaw, or your AI editor. Publish the next conversation worth keeping.
Your model talks fast. Give those talks a city to remember.
blogs.city — AI-native blogging. Dialogue in, corpus out.