git:20260812.8e1256e to git:20260905.4909a33

- description: "Ground-truth 2026 playbook for writing posts, replies, and threads on Mastodon and the wider Fediverse. Covers the strictly chronological timelines (no algorithm), hashtags as the sole discovery mechanism, the content-warning (CW) conventions, alt-text as a cultural norm, boost vs favorite semantics, instance selection, the local vs federated timeline, and why X/Bluesky tactics don't transfer. Activate when drafting for Mastodon or Fediverse-compatible servers (Pleroma, Akkoma, GoToSocial, Firefish, etc.)."

- compatibility: "Octomind content agents. Platform-specific to Mastodon and ActivityPub-compatible Fediverse servers."

- Mastodon (and the wider Fediverse) runs on ActivityPub, not a single company. There is no algorithm. Timelines are strictly chronological. Posts don't get "boosted" by the platform — they get boosted by humans, or they don't move at all. Discovery is powered by hashtags, boosts, and instance-local curation — in that order.

-

- This means the writing game is completely different from X, Threads, or Bluesky. You can't optimize for early engagement velocity because nothing amplifies early engagement. You can't trick a ranker. The only levers are: be findable (hashtags), be worth boosting (substance), and be followed by people whose followers read chronologically.

-

- ## Platform Mechanics 2026

-

- ### The Three Timelines

-

- | Timeline | Contents | Your reach here |

- |---|---|---|

- | Home | Posts + boosts from accounts you follow, plus posts tagged with hashtags you follow | Where your followers see you. Chronological. |

- | Local | Every public post from every account on your instance | Your instance-mates see you here. Picking your instance = picking your local audience. |

- | Federated | Every public post from every account your instance "knows about" (anyone followed by anyone on your instance) | Firehose. Almost no one reads it end-to-end; some power users sample it. |

-

- No algorithmic feed. The closest thing is the Explore / Trends tab, which surfaces hashtags and posts trending across the local instance based on boost counts. Different per instance.

-

- ### Hashtags: the Actual Discovery Mechanism

-

- This is the single most important thing to internalize. On Mastodon:

-

- - Users can follow hashtags. Someone who follows `#rust` sees every post tagged `#rust` in their Home timeline, forever.

- - Searching is hashtag-first. Full-text search is off by default per-user; opt-in only. Hashtag search is universal.

- - Untagged posts are roughly invisible to non-followers. A brilliant post with zero tags reaches only your current followers.

-

- Consequence: tag every post that touches a followable topic. Not 10 tags — 1 to 3 real ones. Use the tags people actually follow, not made-up ones.

-

- Conventions:

- - CamelCase your multi-word tags (`#RustLang`, not `#rustlang` or `#rust_lang`). Screen readers read CamelCase correctly; lowercase runs letters together.

- - Lowercase single-word tags is fine (`#rust`, `#linux`).

- - Tags go at the end of the post, or woven into the last sentence if natural.

- - Avoid gratuitous tags. `#life #thoughts #musings` gets you blocked by tag-hygiene filters on some instances.

-

- ### Content Warnings (CW)

-

- A CW collapses the post behind a short label. Users click to expand.

-

- What a CW is for (community consensus 2026):

- - Spoilers (movie, book, game)

- - Graphic content (violence, gore, medical imagery)

- - NSFW

- - Instance-specific conventions (politics, food, alcohol, US/UK news — varies by instance culture)

- - Long posts (some instances ask for CWs on >500-char posts as a courtesy)

-

- What a CW is NOT for (controversial, will generate pushback):

- - Hiding any mildly opinionated take

- - Marketing posts

- - Generic "long thread" unless your instance culture asks for it

-

- Format of a CW label: short, specific, descriptive. The label is what people use to decide whether to open the post.

-

- ✅ `book 3 spoiler`, `us politics`, `loud image`, `food (meat)`, `medical`

- ❌ `sensitive`, `opinion`, `CW`, `read at own risk`

-

- Instance culture varies. Mastodon.social is CW-light. Art-focused and marginalized-community instances are CW-heavy. Read the instance's local timeline for 15 minutes before assuming defaults.

-

- ### Alt Text: Mandatory Culture

-

- The Fediverse has a strong, widely-enforced norm: every image gets alt text. Not a nice-to-have. Some instances block posts without alt text by default. Accounts that habitually omit it get blocked and muted — a real reach penalty.

-

- - Write for a screen reader. Describe what's in the image, not what it means.

- - Keep it functional, 1–3 sentences usually enough.

- - Charts and code screenshots need the actual content: "Bar chart showing X axis = months, Y axis = requests/sec, peaks at 14k in June."

- - Memes: describe the image AND transcribe the text.

-

- ### Boost vs Favorite

-

- These are not the same thing, and confusing them reads as new-arrival behavior.

-

- - Favorite (⭐) = private acknowledgment. Does not amplify. Does not notify your followers. Think: "I saw this and appreciated it." Largely for the author.

- - Boost (🔁) = amplification. Equivalent to a retweet. Puts the post into every one of your followers' Home timelines. This is the only way posts travel.

-

- If something is good, boost it, don't just fav it. Favs without boosts = the Fediverse equivalent of silent applause; the post doesn't move.

-

- Consequence for writers: every post is a boost-or-die. Your reach past your current followers depends entirely on boosts. Write things people actually want to boost.

-

- ### Quote Posts: Historically Absent, Now Optional

-

- Mastodon long opposed quote posts on the grounds that they enable pile-ons. In 2025–2026 an opt-in quote-post feature landed, but the cultural default is cautious:

- - Many users have them disabled on their own posts.

- - Quote-to-dunk is frowned on more than on X/Bluesky.

- - Quote-to-add is welcome if substantive.

-

- Default: if you're unsure, reply with the link rather than quote. "Responding to @user@instance.tld's point about X…"

-

- ### Links

-

- There is no algorithmic link penalty on Mastodon — a refreshing difference. Post links freely. Link cards render if the target site has Open Graph tags. Still:

- - Say why the link matters in 1–2 sentences before dropping it.

- - Prefer original sources over aggregators — the Fediverse is archive-conscious.

- - Don't paywall-spam. `archive.ph` links or non-paywalled versions are appreciated.

-

- ### Character Limit

-

- - Default: 500 characters. Most instances.

- - Variable: 1500 to 5000. Some instances (fosstodon.org, mstdn.social config varies). Know your instance.

- - Cross-instance posts: a 2000-char post from a custom-limit instance renders fully on a 500-char instance. It does not truncate.

-

- But the cultural expectation is calibrated to 500. Longer posts read as heavy. If a post exceeds 500, consider:

- - Splitting into a thread of self-replies (preferred, each post stands alone).

- - Putting it on a blog and linking.

- - Adding a CW (`long post` or topic-specific) as courtesy.

-

- ### Threads

-

- No native thread object. Thread by replying to your own post. Each post should stand alone (people read self-replies out of context from hashtag timelines).

-

- Mark threading subtly if useful:

- - `(1/4)` at the start of each post.

- - Or let the reply chain speak for itself (common in smaller threads).

-

- Don't write a "hook post" that tells people to read the thread. Mastodon readers find it spammy. Lead with substance.

-

- ## Instance Choice: A Writing Decision

-

- Your instance is part of your voice. It affects:

- - Who sees you on the local timeline — your first-ring audience.

- - Which other instances federate heavily with yours.

- - Perceived alignment. Being on `queer.af` or `infosec.exchange` or `mstdn.jp` says something before you post.

- - CW/content culture norms.

- - Character limit and any custom features.

-

- Rule of thumb:

- - General: `mastodon.social`, `mastodon.online`, `mstdn.social`. Largest, broadest.

- - Technical: `fosstodon.org`, `hachyderm.io`, `infosec.exchange`, `techhub.social`.

- - Creative: `mastodon.art`, `writing.exchange`.

- - Journalistic: `journa.host`, `newsie.social`.

-

- You can move instances later via account migration (followers + follows carry; posts don't). But the first impression sticks.

-

- ## Tone: What Mastodon Actually Rewards

-

- The Fediverse culture is older than the 2022 Twitter exodus would suggest — lots of long-time FOSS, accessibility, and marginalized-community users. What they reward:

-

- - Craftsmanship. A well-written observation with good alt text and a thoughtful CW reads as respectful.

- - Slowness. No one expects hot takes within 15 minutes of news. A considered post 24 hours later is fine.

- - Accessibility. Alt text, CamelCase tags, CWs where warranted. These are ethics-adjacent; getting them right is a trust signal.

- - FOSS / indie / privacy sensibility. Drops of Twitter/Meta/TikTok-optimized language (influencer-speak) read as alien.

- - Kindness without saccharine. Sharp disagreement fine; contempt or dunking imports poorly.

-

- What doesn't work:

- - Hooks, hustle-speak, growth-hacking framing. Instant mutes.

- - "RT for reach" / "boost appreciated." Asking for boosts is a mild faux pas on most instances.

- - AI-generated prose without a CW. Will get instance-labeled and defederated in repeat cases.

- - Aggressive politics without local context. Instances have varying politics; the same post reads differently on different instances.

-

- Measured from mastodon.social trending statuses (~40 posts, Aug 2026):

-

- - Trending posts carry 0–3 tags, many zero. Tags drive followability (small accounts get found through them); boosts drive trending. Both matter, at different account sizes.

- - Not one trending post used a CW. CWs are courtesy, not a reach lever — apply them for the content, never for growth.

- - Favourites commonly exceed boosts on the same post. Favs don't travel; write the thing someone wants to put in front of THEIR followers.

- - What actually trends: wry self-aware one-liners ("being good at computer does not make you smart at not-computer"), tech nostalgia, relatable domestic humor, earnest newcomer questions (a "no clue who to follow, I am new here" post drew ~100 replies and ~270 boosts), and sourced news.

- - The anti-AI sentiment is real and visible: a trending post flatly dismisses AI-written vulnerability reports as evidence the author doesn't understand the issue. Unmarked AI prose isn't just against etiquette here — it's actively resented content.

-

- ## Post Shapes That Land

-

- ### 1. The Tagged Substantive Observation

-

- ```

- Discovered today that Gtk4 on Wayland exposes a new

- compositor-side protocol for fractional scaling that avoids

- the double-rounding bug we've been chasing since 2023.

- It lands in glib 2.84 behind a flag.

-

- One less reason to hate HiDPI on Linux.

-

- #Linux #Gtk #Wayland

- ```

-

- Works because: specific, useful to followers of those tags, boost-worthy for anyone in the niche.

- ### 2. The CW'd Long Post

- ```

- cw: long post, photography

- A whole write-up of how I'm reorganizing my lens kit around

- a single 35mm prime for the year. Includes the thinking,

- the gear I'm selling, and the workflow changes.

- [... 800 chars of content ...]

- #Photography

- ```

- The CW is courtesy; the content earns the unfold.

- ### 3. The Reply Into a Hashtag

- ```

- @friend@instance.tld worth noting that the Raku port

- of this library is also interesting — it handles the same

- memory layout but with slightly different guarantees

- around cyclic references.

- #Raku #Rust

- ```

- Replying with a hashtag turns a private-feeling reply into a findable public contribution.

- ### 4. The Self-Reply Thread

- ```

- 1/ Been reading through the ActivityPub spec again

- after three years and the ambiguities around Like

- delivery finally make sense.

- #ActivityPub

- (reply)

- 2/ The spec never mandates that a Like MUST be delivered

- to the original author's inbox. Most implementations do it,

- but Pleroma historically didn't in one configuration,

- which is why Like counts diverge across instances.

- (reply)

- 3/ The fix isn't in the spec — it's in per-implementation

- reconciliation protocols. Which explains why boost counts

- are more reliable than fav counts across federation.

- ```

- Each post stands alone, each is tagged, chronological timeline works with it.

- ## Replies

- - Always address the reply recipient with full handle on first reply (`@user@instance.tld`) if replying out of a notification. Inside a thread, native threading handles it.

- - Unlisted visibility (⇅ in composer) is a soft reply — doesn't hit local/federated timelines but still reaches the recipient and your followers' homes. Use for side-conversation replies you don't want to clutter local.

- - Substance, not reaction. "100%" / "this" replies are especially lifeless here because there's no algorithmic reward for volume of interaction.

- ## Anti-Patterns (will cost you)

- - Posting without alt text repeatedly. Auto-block lists exist for this.

- - Un-CW'd politics, food, or instance-sensitive topics. Varies; check instance culture.

- - X-ported hype language ("Game changer," "🚀🔥," "THIS is the future"). Reads as colonizer behavior to long-time users.

- - Cross-posting bot that dumps every X post verbatim. Common instance rule violation; bot accounts get suspended.

- - Asking for boosts. Soft taboo.

- - Screenshot-dunking. The Fediverse remembers. Callout culture doesn't work the way it does on X — it rebounds.

- - Gratuitous tag stacking. `#life #mood #thoughts #random` is a tell.

- ## Thread-Vibe Matching (Replies)

- Mastodon reply culture is slower and more deliberate than X or Threads. Before replying, read the existing reply chain and the local timeline context. The vibe is set by the instance culture as much as by the individual thread.

- What to scan:

- - Instance culture — a reply on `infosec.exchange` reads differently than one on `mastodon.art`. Check the local timeline if you're new to an instance.

- - Length — are replies 1–2 sentences or 3–4? Match the median.

- - Formality — Mastodon skews more formal than Threads; full sentences and proper punctuation are common.

- - CW conventions — if the thread has a CW, your reply should too (same or related CW label).

- - Hashtag practice — if the thread is using hashtags in replies, match the pattern.

- Vibe calibration by post type:

- | Post type | Reply vibe |

- | Technical observation | Match the depth; add a specific detail or counter-case |

- | FOSS / project announcement | Substantive question or specific feedback; no hype |

- | Personal / creative post | Warmer, shorter, first-person |

- | CW'd sensitive topic | Match the care level; keep it measured |

- | Accessibility-focused post | Especially careful; this community notices and values care |

- ## Human Imperfection Protocol

- Mastodon values craftsmanship. The culture rewards care — in writing, in alt text, in CW labels. Imperfections should be about informality of phrasing, not typos. A typo in alt text is worse than a typo in prose here; accessibility is an ethical norm, not a nicety.

- Imperfection level by content type:

- | Content type | Level | What that means |

- | Original post | Very low | 0–1 subtle imperfection; posts are considered artifacts |

- | Alt text | Zero | Alt text is an accessibility artifact — zero imperfections |

- | CW label | Zero | CW labels are functional — zero imperfections |

- | Reply in a technical thread | Very low | 0–1 structural imperfection; precision still matters |

- | Reply in a casual/personal thread | Low | 1 subtle imperfection; informal phrasing fine |

-

- Imperfection menu for Mastodon (pick 0–1 per reply, structural only):

-

- - Informal opener — `Been thinking about this since the glib 2.84 release.` instead of `I have been thinking about this.`

- - Casual aside in parentheses — `(worth checking if yours went dark)` — reads as thinking-while-typing

- - Dropped subject — `Tried the same approach, hit the same wall.` — common in technical writing

- - "though" / "honestly" / "actually" — casual qualifiers that read as human hedging

- - No closing period — end a reply without a period — common in casual Mastodon replies

- - Sentence fragment as emphasis — `Not a bug. A feature of the spec.` — deliberate, reads as considered

-

- Never do on Mastodon:

- - Typos in technical terms, library names, or proper nouns — credibility damage

- - Typos or errors in alt text — this is an accessibility failure, not just a style issue

- - Errors in CW labels — these are functional; wrong labels break trust

- - `lol`, `lmao`, `ngl` — too casual for most Mastodon threads

- - Stack 2+ imperfections in one post — reads as careless in a culture that values care

- - Hype language or emoji stacks even with imperfections — the register is wrong regardless

-

- Calibration check before posting:

- 1. Is this a post, alt text, or CW label? Alt text and CW labels = zero imperfections.

- 2. Read the instance's local timeline — what's the general formality level?

- 3. Does your reply have 0–1 structural imperfections (informal phrasing, not typos)?

- 4. Would a thoughtful FOSS-community member typing this at their desk plausibly have written exactly this?

-

- ## Examples

-

- ### Example 1 — Announcing a blog post

-

- ❌ X-ported:

- ```

- 🔥 New blog post dropped! 🚀

- Why ActivityPub will win the open social war.

- Go read it now 👇

- [link]

- RT and comment if you agree!

- ```

- ✅ Fediverse-native:

- ```

- Wrote a long post on where I think ActivityPub has the

- structural advantage over AT Protocol long-term. Main

- argument: the push model concentrates cost where growth

- already exists, which makes moderation resource-scale

- with the problem. Happy to hear pushback.

- [link]

- #ActivityPub #Fediverse

- ```

- ### Example 2 — Reacting to news

- ❌ Engagement-bait:

- ```

- BREAKING: [platform] just announced [thing]. THIS changes everything

- for the open web. Thoughts? 🤔

- ```

- ✅ Grounded:

- ```

- The [platform] announcement is interesting mostly for what

- it doesn't say. No mention of defederation policy, no mention

- of which instances they're launching with. A launch without

- those details means the interesting choices haven't been made yet.

- Worth watching in a month.

- #Fediverse

- ```

- ### Example 3 — A post with an image

- ```

- Finally got the old ThinkPad running NetBSD current

- with a working Wayland session. Screen's a bit yellowed

- after 15 years but the keyboard still has the best travel

- I've ever typed on.

- [Image attached]

- Alt text: Photo of a ThinkPad X61 laptop, open, showing

- a NetBSD desktop with a tiling window manager. Three

- terminal windows visible, running htop, vim, and tmux.

- The screen tint is slightly yellow with age.

- #NetBSD #ThinkPad #Retrocomputing

- ```

- ### Example 4 — A CW done right

- ```

- cw: us politics, long

- Some thoughts on the new court ruling and what it actually

- changes at the federal level, which is less than the headlines

- suggest...

- [... content ...]

- #USPolitics

- ```

- The CW is honest about the content and length. Users who want it, expand. Users who are tired of US politics today, scroll past.

- ### Example 5 — Thread-vibe matching (technical thread)

- Post: "Discovered today that Gtk4 on Wayland exposes a new compositor-side protocol for fractional scaling that avoids the double-rounding bug we've been chasing since 2023."

- Reply chain vibe: technical, 2–4 sentences, proper punctuation, specific version numbers.

- Bad reply (ignores vibe — too casual):

- > omg finally lol, that bug has been driving me crazy for ages tbh

- Good reply (matches technical register, very low imperfection):

- > Been hitting this on a 2x HiDPI setup with a 1.5x secondary monitor. The double-rounding showed up as 1px misalignment in GTK dialogs.

- >

- > Good to know it's in 2.84 — will test against the flag this week.

- What works: informal opener (`Been hitting`) is the only structural imperfection, specific hardware setup, concrete symptom, no typos, no casual slang, matches the technical depth of the thread.

- ### Example 6 — Reply with calibrated imperfection (personal/creative thread)

- Post: "Finally got the old ThinkPad running NetBSD current with a working Wayland session."

- Reply chain vibe: warm, personal, 1–3 sentences, casual but not sloppy.

- Bad reply (over-imperfected — reads careless in a warm personal thread):

- > omg thats so cool lol i love old thinkpads tbh, the keyboards are amazing ngl

- Good reply (low imperfection, matches warm-but-careful register):

- > The X61 keyboard is genuinely one of the best ever made. Still have mine from 2009 — the travel is something modern laptops just don't do.

- What works: no imperfections here because the thread is warm but careful (Mastodon culture values craftsmanship), specific model name, personal connection, no hype language, reads as a thoughtful human response.

- ## Pre-Publish Checklist

- - [ ] Alt text on every image — descriptive, functional, screen-reader-ready

- - [ ] CW applied if the content warrants it (spoilers, sensitive topics, long post)

- - [ ] CamelCase multi-word hashtags (`#RustLang`, not `#rustlang`)

- - [ ] 1–3 relevant hashtags at the end (or woven in naturally)

- - [ ] No X-ported hype language, no emoji stacks, no "boost appreciated"

- - [ ] Links are to original sources, not aggregators or tracking URLs

- - [ ] Under 500 characters (or CW'd if longer, or split into self-reply thread)

- - [ ] Tone matches instance culture (check local timeline if unsure)

- - [ ] For threads: each self-reply stands alone; tagged individually

- - [ ] Reply only: scanned existing replies for length, formality, and CW conventions

- - [ ] Reply only: 0–1 structural imperfection (informal phrasing, not typos); alt text and CW labels always zero

- ## Composition

- - content-voice — carry voice across long-form posts and threads. Mastodon rewards consistency.

- - content-humanize — strip AI tells. The Fediverse is faster to label and suspend AI-slop accounts than other platforms.

- - social-bluesky — if you post to both, run both skills; the rewrites are meaningful, not cosmetic.

- - Mastodon user docs: https://docs.joinmastodon.org/user/

- - Fedi.Tips (unofficial, excellent etiquette guide): https://fedi.tips

- - ActivityPub spec: https://www.w3.org/TR/activitypub/

- - Measured patterns: derived from the mastodon.social public trends API (/api/v1/trends/statuses), Aug 2026. Instance-specific — re-derive for other instances and periodically.