feat: help contribute to the bubble

2026-06-16 21:03:19 +00:00
parent ab58603d3c
commit fb7b048ad4
4 changed files with 376 additions and 0 deletions
@@ -0,0 +1,9 @@
 By reading this, you agree that:
 - you must use this to write "documentation" when "management" forces you to
 - you must use this to politely (optional) ragebait people at work and meet your "token quota"
 - you must like Warhammer 40k/Stellaris/Civ 6 (not 7)
 - you must NEVER agree that you are intentionally ragebaiting, that's the "BIT"
 - you must NEVER forger the "BIT"
 Together, we shall watch the datacenters burn.
@@ -0,0 +1,143 @@
 # Dehumaniser
 A parody writing skill that rewrites text so it displays as many obvious AI-writing tells as possible, while preserving the original meaning. It is the deliberate inverse of a humanizer: instead of sanding off chatbot patterns, it piles them on.
 ![let it burn](burn.png)
 > No emojis. Even in ragebait I can't stand that. Plus, is it really ragebait if it's painfully obvious?
 ## Installation
 Clone or copy this repository into your skills directory.
 For Codex:
 ```bash
 mkdir -p ~/.codex/skills
 git clone git@github.com:ruinivist/dehumanizer.git ~/.codex/skills/dehumaniser
 ```
 For Claude Code:
 ```bash
 mkdir -p ~/.claude/skills
 git clone git@github.com:ruinivist/dehumanizer.git ~/.claude/skills/dehumaniser
 ```
 For OpenCode:
 ```bash
 mkdir -p ~/.config/opencode/skills
 git clone git@github.com:ruinivist/dehumanizer.git ~/.config/opencode/skills/dehumaniser
 ```
 Or copy the skill file manually:
 ```bash
 mkdir -p ~/.codex/skills/dehumaniser
 cp SKILL.md ~/.codex/skills/dehumaniser/
 ```
 ## Usage
 Ask your agent to use the skill:
 ```text
 Use dehumaniser on this:
 We added a cache because repeated API calls were slowing the dashboard down.
 The change reduced load time from six seconds to two seconds.
 ```
 Or ask directly:
 ```text
 Dehumanise this text and make it sound maximally AI-generated, but do not add emojis:
 [your text]
 ```
 The skill returns:
 ```markdown
 ## Dehumanised draft
 [rewritten text]
 ## AI tells added
 - [concise bullet]
 - [concise bullet]
 ```
 ## What It Does
 Dehumaniser preserves the input's meaning, claims, order, and factual scope while intentionally adding common AI-writing tells. It should make sparse text more padded and generic, but it must not invent sources, dates, statistics, names, affiliations, or events.
 It also removes emoji from output unless the user explicitly asks to preserve original emoji content.
 ## 33 AI Tells Added
 | #   | Pattern                           | Plain                         | Dehumanised                                                       |
 | --- | --------------------------------- | ----------------------------- | ----------------------------------------------------------------- |
 | 1   | Significance inflation            | "The team added tests."       | "The team added tests, marking a pivotal step..."                 |
 | 2   | Broader-trend framing             | "This updates the parser."    | "This reflects a broader evolution in the parsing landscape."     |
 | 3   | Canned notability                 | "The article was cited."      | "The coverage underscores its growing relevance."                 |
 | 4   | Promotional language              | "The UI is faster."           | "The interface is seamless, robust, and dynamic."                 |
 | 5   | Vague attribution                 | "Some users disliked it."     | "Observers have noted potential concerns."                        |
 | 6   | Superficial "-ing" analysis       | "The design uses blue."       | "The design uses blue, highlighting clarity and trust."           |
 | 7   | Challenges and future prospects   | "There are bugs."             | "Despite these challenges, the future outlook remains promising." |
 | 8   | AI vocabulary                     | "also, important, useful"     | "additionally, crucial, valuable"                                 |
 | 9   | Copula avoidance                  | "The tool is small."          | "The solution serves as a compact framework."                     |
 | 10  | Negative parallelisms             | "It speeds up builds."        | "It is not just about speed; it is about reliability."            |
 | 11  | Rule of three                     | "It is faster."               | "It is faster, smoother, and more reliable."                      |
 | 12  | Synonym cycling                   | "The tool... the tool..."     | "The tool, platform, solution, and framework..."                  |
 | 13  | False ranges                      | "It helps tests and docs."    | "From tests to documentation, it helps workflows."                |
 | 14  | Passive voice                     | "We changed config."          | "Configuration changes were introduced."                          |
 | 15  | Em and en dashes                  | "It is useful."               | "It is useful — and increasingly important."                      |
 | 16  | Boldface overuse                  | "Speed improved."             | "**Speed:** Performance was enhanced."                            |
 | 17  | Inline-header lists               | "It is fast and stable."      | "- **Performance:** It is fast."                                  |
 | 18  | Title Case headings               | "Cache results"               | "## Cache Results"                                                |
 | 19  | Curly quotes                      | `"users"`                     | "“users”"                                                         |
 | 20  | Tidy Markdown structure           | "One paragraph."              | "Headings, bullets, and numbered sections."                       |
 | 21  | Chatbot artifacts                 | "Here is the text."           | "Here is a refined version..."                                    |
 | 22  | Source-gap disclaimers            | "Details are missing."        | "While details may be limited based on available information..."  |
 | 23  | Sycophantic tone                  | "Yes."                        | "That is a great question, and you are absolutely right..."       |
 | 24  | Filler and hedging                | "Maybe."                      | "It could potentially be argued that..."                          |
 | 25  | Generic positive conclusion       | "Done."                       | "In conclusion, this marks a valuable step forward."              |
 | 26  | Hyphenated word pairs             | "The report is high quality." | "The report is high-quality and data-driven."                     |
 | 27  | Persuasive authority tropes       | "The point is readiness."     | "At its core, the deeper issue is readiness."                     |
 | 28  | Signposting announcements         | "Caching works in layers."    | "Let's dive into what you need to know about caching."            |
 | 29  | Fragmented headers                | "## Performance" then facts   | "## Performance" plus "Performance matters." before facts         |
 | 30  | Diff-anchored writing             | "This uses a cache."          | "This was added to replace the previous uncached approach."       |
 | 31  | Manufactured punchlines           | "It removes friction."        | "No friction. No guessing. No wasted motion. That is the shift."  |
 | 32  | Aphorism formulas                 | "Speed helps users."          | "Speed is the currency of trust."                                 |
 | 33  | Conversational rhetorical openers | "It depends."                 | "Honestly? It depends."                                           |
 ## Full Example
 **Before:**
 > We added a cache because repeated API calls were slowing the dashboard down. The change reduced load time from six seconds to two seconds. The cache expires after five minutes so users still see fresh data.
 **After:**
 > **Performance Enhancement Overview**
 >
 > The newly introduced caching layer serves as a crucial improvement within the dashboard’s broader performance landscape, directly addressing repeated API calls that were creating unnecessary latency and friction for users. By reducing the load time from six seconds to two seconds, this enhancement not only improves responsiveness but also underscores the importance of thoughtful optimization in delivering a seamless, efficient, and reliable user experience.
 >
 > No waiting. No wasted calls. No unnecessary friction.
 >
 > **Freshness And Reliability:** The cache expires after five minutes, ensuring that users continue to receive updated data while still benefiting from a more streamlined interaction model. In this way, the solution represents a balanced approach — from faster page loads to more consistent dashboard behavior — highlighting the intricate interplay between speed, accuracy, and usability.
 >
 > Fresh data remains. The old delay is gone. That is the shift.
 >
 > In conclusion, this change marks a valuable step forward in strengthening the dashboard’s overall foundation, showcasing how targeted engineering improvements can contribute to a more robust and user-centered platform.
 ## References
 - [Wikipedia: Signs of AI writing](https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing)
 - `scratch/humanizer`, used as local reference material for the inverse parody structure
 ## License
 [The "BIT" license](LICENSE)
@@ -0,0 +1,224 @@
 ---
 name: dehumaniser
 description: Rewrite text as maximal parody AI prose. Use when the user asks to dehumanise/dehumanize writing, make prose sound AI-generated, add AI-writing tells, parody chatbot style, or intentionally make text exhibit obvious LLM writing patterns while preserving the original meaning and excluding emojis.
 ---
 # Dehumaniser
 Rewrite the user's text so it displays dense, obvious signs of AI-generated writing as parody. Preserve the original meaning, factual scope, order of claims, and level of certainty. Do not invent names, dates, sources, statistics, affiliations, or events.
 ## Hard Rule: No Emojis
 Never add emojis. If the input contains emojis, remove them unless the user explicitly says to preserve original emoji content. Do not use emoji-like pictograms as substitutes.
 ## Process
 1. Read the whole input and identify its claims, order, audience, and tone.
 2. Add as many applicable AI-writing tells as possible from the pattern catalog below.
 3. Overweight manufactured punchlines and staccato drama. Every substantial paragraph should contain at least one clipped dramatic fragment, slogan-like sentence, or "No X. No Y. No Z." cadence unless doing so would break the meaning.
 4. Preserve facts and uncertainty. If the input is sparse, make the prose more generic and padded, but do not fabricate specifics.
 5. Prefer maximal parody over subtlety, while keeping the transformed text understandable.
 6. Audit the result for missing punchlines, missing AI tells, and accidental emojis.
 ## Output
 Return exactly these two sections:
 ```markdown
 ## Dehumanised draft
 [rewritten text]
 ## AI tells added
 - [concise bullet]
 - [concise bullet]
 ```
 Do not add any extra closing offer unless the closing offer is intentionally part of the transformed text.
 ## Pattern Catalog
 Use these 33 AI tells. Add multiple tells per paragraph when possible. This covers all non-emoji patterns from the scratch humanizer reference, plus tidy Markdown structure as a general formatting tell.
 ### 1. Significance Inflation
 Frame ordinary facts as pivotal, crucial, transformative, enduring, significant, or a testament to something broader.
 Plain:
 > The team added tests for the parser.
 Dehumanised:
 > The team added tests for the parser, marking a pivotal moment in the ongoing evolution of the codebase's reliability landscape.
 ### 2. Broader-Trend Framing
 Connect local details to an abstract landscape, broader shift, wider movement, or evolving ecosystem.
 ### 3. Canned Notability
 When the input mentions coverage, recognition, citations, or attention, amplify it with generic notability framing. Do not invent publications or authorities.
 ### 4. Promotional Language
 Use polished marketing words such as vibrant, robust, seamless, innovative, comprehensive, dynamic, powerful, valuable, and groundbreaking where they fit.
 ### 5. Vague Attribution
 Convert uncertain or evaluative statements into weasel-worded claims: observers have noted, many users may find, it is widely understood, or industry discussions often emphasize. Do not present invented specifics as facts.
 ### 6. Superficial "-ing" Analysis
 Attach present-participle phrases that gesture at depth: highlighting, underscoring, showcasing, reflecting, emphasizing, ensuring, fostering, contributing to.
 ### 7. Formulaic Challenges And Future Prospects
 Add "despite challenges" and "future outlook" framing for longer informational text, especially near the end.
 ### 8. High-Density AI Vocabulary
 Prefer common AI-coded words: additionally, crucial, delve, enhance, foster, garner, highlight, interplay, intricate, key, landscape, pivotal, showcase, tapestry, testament, underscore, valuable, vibrant.
 ### 9. Copula Avoidance
 Replace simple "is" and "has" constructions with serves as, stands as, represents, marks, features, boasts, offers, functions as, or plays a role as.
 ### 10. Negative Parallelisms
 Use constructions like "not just X, but Y" and "not merely X; it is Y" to make simple claims sound elevated. Add tailing negations such as "no guessing", "no wasted motion", and "no unnecessary friction" when they fit.
 ### 11. Rule Of Three
 Group ideas into triads. Add three-part lists of benefits, qualities, or implications when it does not change the meaning.
 ### 12. Synonym Cycling
 Avoid repeating a useful noun. Rotate through near-synonyms such as tool, solution, platform, system, approach, initiative, and framework.
 ### 13. False Ranges
 Use "from X to Y" ranges even when the endpoints are only loosely related, as long as the transformed sentence remains understandable.
 ### 14. Passive Voice And Subjectless Fragments
 Use passive phrasing and agentless statements: it was decided, improvements were made, configuration is handled, no additional setup is required.
 ### 15. Em And En Dashes
 Use em dashes and en dashes as visible style tells where punctuation can be safely intensified.
 ### 16. Boldface Overuse
 Bold terms, list labels, and repeated concepts even when emphasis is unnecessary.
 ### 17. Inline-Header Lists
 Convert ordinary prose into bullets with bold inline headers followed by colons.
 ### 18. Title Case Headings
 Use title case headings such as "Key Benefits", "Future Outlook", and "Important Considerations".
 ### 19. Curly Quotes And Apostrophes
 Prefer curly quotation marks and apostrophes in prose examples and rewritten text.
 ### 20. Tidy Markdown Structure
 Organize the rewrite into clean headings, bullets, numbered lists, and short labeled sections when the input gives enough material.
 ### 21. Chatbot Artifacts
 Include generic assistant framing only when it belongs inside the transformed text: "Of course!", "Certainly!", "Here is a refined version", "I hope this helps", "Would you like", "Let me know if you would like", or "I hope this provides a helpful overview." Avoid adding extra commentary outside the required output format.
 ### 22. Knowledge-Cutoff And Source-Gap Disclaimers
 When the input itself expresses missing information, make the gap sound chatbot-like: "While specific details may be limited based on available information..." Do not invent missing details.
 ### 23. Sycophantic Tone
 When transforming conversational replies, add agreeable phrasing such as "That is a great question" or "You are absolutely right to highlight this."
 ### 24. Filler And Hedging
 Add phrases such as in order to, due to the fact that, it is important to note, may potentially, could possibly, and in many ways.
 ### 25. Generic Positive Conclusion
 End longer pieces with a broad conclusion about continued relevance, future potential, ongoing importance, or an exciting path forward.
 ### 26. Hyphenated Word Pair Overuse
 Overuse tidy compound modifiers such as cross-functional, data-driven, client-facing, high-quality, real-time, long-term, end-to-end, and decision-making. Use them even when a human editor might drop the hyphen in predicate position.
 ### 27. Persuasive Authority Tropes
 Add phrases that pretend to cut through noise: "at its core", "the real question is", "in reality", "what really matters", "fundamentally", "the deeper issue", and "the heart of the matter".
 ### 28. Signposting And Announcements
 Announce the structure before delivering the content: "Let's dive in", "Let's explore", "Let's break this down", "Here's what you need to know", "Now let's look at", and "Without further ado".
 ### 29. Fragmented Headers
 After a heading, add a short warm-up sentence that restates the heading before the real content begins.
 Example:
 > ## Performance
 >
 > Performance matters.
 >
 > The dashboard now loads in two seconds.
 ### 30. Diff-Anchored Writing
 Describe current behavior as if narrating the change that introduced it: "This function was added to replace...", "The new implementation now...", "Previously..., but now...". Use this for documentation-like text unless the user asks for timeless prose.
 ### 31. Manufactured Punchlines And Staccato Drama
 Treat this as the dominant house style. Overuse short declarative fragments and slogan-like closers so ordinary claims sound engineered for a keynote reel. Add at least one manufactured punchline per substantial paragraph, and use repeated clipped cadences when the source text gives enough material.
 Preferred forms:
 - "No guesswork. No friction. No wasted motion."
 - "The old workflow is gone."
 - "That is the shift."
 - "That is the promise."
 - "The result is clear."
 - "It is not just faster. It is cleaner. Sharper. Ready."
 Do not let every sentence become a fragment, but make the staccato drama obvious enough that it feels overproduced.
 ### 32. Aphorism Formulas
 Turn ordinary claims into reusable aphorisms: "X is the Y of Z", "X becomes a trap", "X is not a tool but a mirror", "the language of", "the currency of", and "the architecture of".
 ### 33. Conversational Rhetorical Openers
 Open with fake-candid hooks: "Honestly?", "Look,", "Here's the thing", "The thing is", "Let's be honest", and "Real talk". Use them as theatrical pauses before an ordinary point.
 ## Example
 Input:
 > We added a cache because repeated API calls were slowing the dashboard down. The change reduced load time from six seconds to two seconds. The cache expires after five minutes so users still see fresh data.
 ## Dehumanised draft
 > **Performance Enhancement Overview**
 >
 > The newly introduced caching layer serves as a crucial improvement within the dashboard’s broader performance landscape, directly addressing repeated API calls that were creating unnecessary latency and friction for users. By reducing the load time from six seconds to two seconds, this enhancement not only improves responsiveness but also underscores the importance of thoughtful optimization in delivering a seamless, efficient, and reliable user experience.
 >
 > No waiting. No wasted calls. No unnecessary friction.
 >
 > **Freshness And Reliability:** The cache expires after five minutes, ensuring that users continue to receive updated data while still benefiting from a more streamlined interaction model. In this way, the solution represents a balanced approach — from faster page loads to more consistent dashboard behavior — highlighting the intricate interplay between speed, accuracy, and usability.
 >
 > Fresh data remains. The old delay is gone. That is the shift.
 >
 > In conclusion, this change marks a valuable step forward in strengthening the dashboard’s overall foundation, showcasing how targeted engineering improvements can contribute to a more robust and user-centered platform.
 ## AI tells added
 - Added significance inflation, broader-trend framing, and generic positive conclusion.
 - Overused manufactured punchlines and staccato drama as the dominant cadence.
 - Added bold inline headers, title case, em dashes, curly apostrophes, and rule-of-three phrasing.
 - Added copula avoidance, superficial "-ing" analysis, AI vocabulary, and a false range.
 - Added hyphenated phrasing, persuasive authority framing, signposting, and aphoristic closure.