Markdown Archives for AI: 2026 Stack with tweet.md + Bluesky

Why this category just showed up in July 2026

If you write long-form on the open social web, you have probably felt the gap: your X Article, your Bluesky long-form post, your LinkedIn newsletter, and your Substack post all need to live as durable Markdown somewhere a future LLM agent can read them. Plain HTML scrapes are messy. Cloud copy-pastes get locked in proprietary editors. What you want is a stack where the source of truth is a .md file, the ingestion is friendly to AI agents, and the publishing layer can target any destination.

Three signals from July 2026 made this an actual category instead of an aspirational hashtag:

• tweet.md started selling credits. The X-to-Markdown converter that launched as a Show HN now ships paid tiers at $5 / $19 / $49 (250 / 1,100 / 3,200 credits) with a free demo corpus of one post, one thread, one X Article, and one profile.
• Bluesky acquired the atproto trademark on July 15, 2026, after another company threatened to enforce it. Bluesky says the move is defensive — everyday use stays free — but the mark is now legally defended for the open ecosystem.
• Toni Schneider became Bluesky's permanent CEO on July 10, 2026, making permanent the role he held on an interim basis since March. Schneider ran Wordpress.com and Automattic's photo and video products before joining Bluesky, which signals a serious go-to-market push.

None of these is a product launch in the marketing sense. All three of them are infrastructure moves that make the LLM-archive stack sturdier. This article is the working draft of that stack.

The three layers of the stack

Think of the stack as three layers you can mix and match:

1. Ingestion — getting other people's content (X threads, X Articles, profiles) into clean Markdown. tweet.md is the canonical 2026 tool here.
2. Publication home — where your long-form Markdown actually lives. Bluesky long-form is the 2026 open-protocol answer; X Articles is the walled-garden alternative; Medium and Substack are the legacy publishing homes.
3. Conversion — turning the canonical Markdown into rich text that survives the destination platform's paste. md2rich runs entirely in the browser and is the conversion layer.

The trick is that the same .md file moves through all three layers. A thread you ingest with tweet.md becomes the source for a Bluesky long-form post, then renders cleanly into LinkedIn via md2rich, then lives in your Obsidian vault as a durable archive. Nothing about the stack is novel in isolation; what is new is that all three layers have shippable 2026 tools.

Ingestion: how tweet.md changed in 2026

tweet.md's premise is one line: replace x.com with tweet.md in any X / Twitter URL and you get back clean Markdown. The free path is a no-charge demo corpus you can fetch any time without an account:

curl "https://tweet.md/*t-twmduser/status/*t-post?thread=off"

The paid path covers your own posts and threads. Pricing as of mid-July 2026:

• $5 — 250 credits (1 credit per post, 2 extra per unique author for metadata)
• $19 — 1,100 credits
• $49 — 3,200 credits

The skill install for AI agents is one line:

npx skills add tweet-md/skill

That command installs the tweet.md skill into Claude Code, Codex, Copilot, Cursor, Gemini CLI, Hermes, OpenCode, and a dozen other coding agents. Once installed, the agent rewrites X URLs in your conversation to tweet.md URLs automatically, and the LLM gets clean Markdown instead of scraping JS-rendered X HTML. That is the piece most people miss: tweet.md is not just a browser tool. It is a backend for AI agents.

Output format is deliberately LLM-friendly. Their docs say the Markdown is optimized for token efficiency, which means short paragraph breaks, no extra spacing around lists, no embedded media placeholders that waste tokens. If you have ever pasted an X URL into ChatGPT and gotten "I cannot access external sites," tweet.md is the workaround that lets the conversation flow.

Publication home: Bluesky's atproto moves in July 2026

Bluesky long-form launched on May 28, 2026, and went from launch curiosity to a serious publication home in seven weeks. Two infrastructure moves in July 2026 are the reason.

First, the atproto trademark. Bluesky PBC announced on July 15, 2026 that it had acquired the rights to the ATPROTOCOL, AT Protocol, and atproto marks from "another company that was threatening to take legal action preventing the company and others from using the term." The post, written by Jim Ray and reviewed by Bluesky's legal team, is explicit: this is defensive, not commercial. "We acquired the rights and are going to defend it from bad actors while making it freely available to the ecosystem." The FAQ spells out that naming an open-source package atproto-feed-tool or AT Protocol SDK does not need a license. Building a paid service whose brand depends on the mark would.

Why this matters for archives: the term "atproto" is now legally stable for the open ecosystem. If you write a Markdown archive that mentions ATProto, atproto, or AT Protocol as the protocol your Bluesky long-form post lives on, that text is not going to be challenged by a trademark troll next year. Bluesky plans to transfer the mark to "an appropriate, independent protocol governance organization" eventually, which is the right endgame.

Second, Toni Schneider as permanent CEO. The Bluesky team blog post on July 10, 2026 makes Schneider's interim role permanent. He has been interim CEO since March. Schneider co-founded and led Wordpress.com and Automattic's photo and video products, and his career arc is the operator's arc: scale, monetization, B2B contracts. For writers, the takeaway is that Bluesky long-form is no longer being run as a research project. The team is being staffed and the platform is being positioned for paid tiers and business contracts in 2027.

That is a good outcome for archives. Your Bluesky long-form Markdown will not be deprecated out from under you in the next product pivot. The platform is being run like a business, which is what durable publishing homes look like.

Conversion: where md2rich actually fits

You have clean Markdown. You want it to look right on LinkedIn, X Articles, Medium, Substack, Notion. The destination editors do not render Markdown on paste. They render rich text, which means you need a converter.

md2rich is a single-page browser tool: a textarea on the left for Markdown, a rendered preview on the right, and a Copy as Rich Text button that puts formatted HTML on your clipboard. Paste that into LinkedIn's post composer and the headings, bold, italic, lists, tables, and code blocks all arrive intact. Paste it into X Articles and you get the same result, modulo X's narrower formatting feature set.

The privacy-first design is the part that matters for archives. md2rich runs entirely client-side; the JavaScript loads once from a CDN and then no network requests happen while you are editing. Your draft, your LinkedIn paste, and your Bluesky long-form text never hit a server. For an archive workflow where you are moving sensitive drafts through a converter, that is the design you want.

In the stack, md2rich is the third layer: it does not compete with tweet.md (which ingests X content) or with Bluesky (which is a publication home). It is the conversion glue that takes whatever Markdown you have at the end of your editing pass and renders it for the destination editor.

A worked example: archiving a thread to LinkedIn

Concretely, here is how the three layers combine for a real workflow. Say you read a long X thread that you want to keep, summarize for your newsletter, and post to LinkedIn as a long-form post.

1. Replace x.com with tweet.md in the thread URL. You get back clean Markdown for the full thread.
2. Open your editor (Obsidian, VS Code, plain text). Trim the Markdown, add your commentary, and save as thread-notes-2026-07-20.md.
3. Paste the trimmed Markdown into md2rich. The preview renders with headings, lists, and links intact.
4. Click Copy as Rich Text and paste into the LinkedIn post composer. The format survives.
5. Cross-post the same .md source to Bluesky long-form. Markdown-friendly editor keeps the structure.

The same .md file is now in your Obsidian vault, on Bluesky, and on LinkedIn. Three surfaces, one source of truth, no cloud lock-in.

Comparison table: tweet.md, Bluesky long-form, and md2rich at a glance

Here is how the three layers compare on the dimensions that matter for a 2026 archive stack. The table is deliberately short; the goal is to pick the right tool for each layer, not to crown a winner.

Privacy: where the data actually lives

For an LLM-archive stack, privacy is the design constraint that decides which tool you trust at each layer.

• tweet.md runs the conversion on its servers. The free demo corpus is hosted by tweet.md and you do not need an account to fetch it. Paid conversions use a cookie-stored API key in your browser; the request body (the X URL) is sent to tweet.md's API and the response (clean Markdown) comes back. There is no on-disk cache by default.
• Bluesky long-form is federated via ATProto. Your post lives in your Personal Data Server (PDS); you can self-host one or use Bluesky's hosted PDS. Exports are documented in the ATProto spec. The platform itself does not train on your posts by default; data use is opt-out per post.
• md2rich is 100% client-side. The JavaScript bundle is fetched once from a CDN; no request leaves your browser after that. Your draft, your preview, and your clipboard payload stay on your device.

The pattern that respects all three layers: use tweet.md to ingest public X content, use Bluesky (or a self-hosted PDS) as the durable home for your own writing, and use md2rich for the final conversion step into destination editors. No single layer holds the full text of your archive.

Who actually needs this stack

Not everyone. Here is the audience that gets real value from a three-layer LLM-archive stack in 2026.

• Newsletter writers who cross-post. If you publish a weekly newsletter and repurpose it to LinkedIn, Bluesky, and Medium, having one .md source with three rendered outputs saves hours of reformatting per week.
• Researchers and analysts. If your job is to summarize X threads, Reddit posts, and long-form articles into briefings, tweet.md is the cheapest way to get clean Markdown into your LLM context window.
• Builders of agent skills. If you ship Claude Code or Codex skills that read social content, the tweet.md skill is one line to add and saves your skill author from scraping X HTML.
• Writers worried about platform lock-in. If you have ever lost a writing surface to a platform shutdown (Twitter's API pricing, Tumblr's adult content ban, Google's various product kills), you already know why a portable .md archive matters.
• Privacy-first creators. If you do not want your drafts sitting on a converter company's server, the md2rich client-side path is the one that respects your threat model.

When this stack is overkill

Honest framing: most writers do not need three tools and an Obsidian vault. You can post to LinkedIn directly without ever touching Markdown, and nobody will penalize you.

The LLM-archive stack is worth the overhead when at least one of these is true:

• You cross-post to three or more platforms regularly.
• You care about keeping a local copy of everything you publish.
• You use AI agents to read or summarize social content.
• You have been bitten by a platform shutdown or API change in the last five years.
• You write long-form (1,000+ words) more than twice a month.

If none of those apply, skip this stack. LinkedIn's native editor is fine for short posts and the marginal value of Markdown portability is zero.

FAQ

Is tweet.md really free in 2026?

Partially. tweet.md ships a no-charge demo corpus (one post, one thread, one X Article, one profile) that you can fetch as Markdown for free. Converting your own posts and threads costs credits: $5 for 250 credits, $19 for 1,100, and $49 for 3,200. One post costs 1 credit; adding author metadata costs 2 more credits per unique author. The no-charge path covers proof-of-concept work but a serious archive run (say, 200 threads you want to summarize) runs into paid territory fast.

What did the Bluesky atproto trademark actually change?

Bluesky PBC acquired the rights to the ATPROTOCOL, AT Protocol, and atproto marks on July 15, 2026, after another company threatened legal action. The acquisition is defensive: Bluesky says it is not charging everyday developers for using the term. Commercial products that use the mark as a brand name (a paid service called "ATProto Cloud", say) would need a license. The takeaway for Markdown-first writers: "atproto" as a publishing infrastructure word is now stable and legally protected for the open ecosystem, which makes Bluesky long-form a less risky home for archives that need a 3-5 year shelf life.

Why did Toni Schneider becoming Bluesky CEO matter for Markdown writers?

Schneider became permanent CEO on July 10, 2026 (he had been interim since March). He previously co-founded and led Wordpress.com and Automattic's video and photo products, and he is the kind of operator investors expect to ship paid tiers and B2B contracts. For writers who keep a Markdown archive of long-form posts on Bluesky, the signal is straightforward: the platform is now being run like a business, not a research project, and the team is being staffed to make long-form and federation stick. That makes your archive less likely to be deprecated out from under you.

How does md2rich fit the LLM-archive stack?

md2rich handles the third leg: you have clean Markdown in (from tweet.md, from your own drafts, from Bluesky long-form copy-paste), and you need it rendered as rich text on a destination platform (LinkedIn, X Articles, Medium, Substack, Notion). md2rich runs entirely in the browser, so the Markdown never leaves your device until you paste it. It complements rather than competes with tweet.md (input side) and Bluesky (publication home) because the conversion step is what makes your archive actually usable across surfaces.

Do I need to be technical to use this stack?

No. tweet.md's value proposition is that you replace x.com with tweet.md in any URL and get clean Markdown back. There is an npx skills add tweet-md/skill command for AI agents, but the manual URL swap works in any browser. md2rich is a single-page browser tool with a textarea on the left and a rendered preview on the right; you paste Markdown, you click Copy as Rich Text, and you paste into LinkedIn or X. Bluesky long-form works like X Articles with a Markdown-friendly editor. The whole stack runs without a terminal.

Will my archived Markdown be usable for AI training later?

That depends on where you archive. tweet.md output is "optimized for LLMs, saving tokens" by their own docs, so it is designed to be cheap input for downstream AI tools. Bluesky is opt-in or opt-out per post via the platform's data settings; the atproto trademark acquisition does not change that. If you want an archive you control, the safest pattern is to keep a local folder of .md files (or an Obsidian vault), pipe source content through tweet.md, and copy-paste rendered output through md2rich. Your Markdown stays portable and human-readable even if every platform changes its API next year.

Conclusion: the stack is real, and the tools ship in 2026

The July 2026 news is not a marketing wave — it is three concrete moves that close gaps in a stack that has been forming for two years. tweet.md has paid tiers so you can ingest your own X content at scale. Bluesky owns the atproto trademark so the protocol's vocabulary is legally stable. Bluesky has a permanent CEO who has shipped consumer products at scale before. md2rich is the conversion layer that turns the canonical Markdown into something every destination editor accepts.

If you write long-form for the open social web, this is the stack to learn. The barrier to entry is low: install the tweet.md skill if you use a coding agent, set up a Bluesky account and write one long-form post, and bookmark md2rich for the conversion step. You will be running the LLM-archive workflow inside an afternoon.