We Are Going to Kill “Unalive” - CourtneyR.dev's 2nd Brain

We Are Going to Kill “Unalive” ![rw-book-cover](https://www.anildash.com/images/broken-monitor.jpg) 🔗 **Source:** [We Are Going to Kill “Unalive”](https://www.anildash.com/2026/10/06/kill-unalive/) > [!note]+ Document Notes > tags: [platform-moderation, algorithmic-censorship, euphemism, content-warnings, open-web] > author: Anil Dash > source: anildash.com > reference: https://www.anildash.com/2026/10/06/kill-unalive/ > date created: 2026-10-07 09:43 > > > [!info] Generated analysis — not the author's words. Sections below are machine-written > > from the captured source. Only text inside `> [!quote]` blocks is verbatim from the author. > > Do not quote or attribute any other line to the named author. > > # 🏷️ Title: We are going to kill “unalive” >  ## 🧭 Table of contents - [📝 Summary](#-summary) - [📖 Main content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning styles assessment](#-learning-styles-assessment) - [📊 Accessibility assessment](#-accessibility-assessment) - [🌍 DEIB assessment](#-deib-assessment) - [✅ Logical validity](#-logical-validity) - [🛡️ DARVO analysis](#️-darvo-analysis) - [🎯 Conclusion](#-conclusion) - [🔄 Refactor appendix](#-refactor-appendix--we-are-going-to-kill-unalive) ## 📝 Summary This is a descriptive reading of a personal essay. It does not give mental-health guidance, and it adds no advice about suicide, self-harm, or sexual violence beyond what the author argues. Anil Dash published the essay on anildash.com on 2026-10-06 (date per the task metadata). It is opinion and advocacy, written in the first person with strong language, and it argues for a change in how people write on platforms. The captured text shows no sponsor, advertisement, or commercial disclosure. - **What happened.** Dash argues that "unalive," misspellings such as r*pe, and emoji substitutes like 🍇 spread because people believe (or have found) that platforms punish plain words for death and sexual assault (P2, P4, P7). Dash says the habit began on algorithmic-feed platforms such as TikTok and now shows up in speech (P3). Dash then calls on readers to say what they mean, build presences on several platforms and on sites they control, and push back publicly when a platform limits them (P16-P30). - **The claim.** The central argument is that euphemism written to appease an algorithm costs people twice. It filters out the people who use content filters (P6), and it trains a generation to let platforms choose their words (P9). The thesis line is the call to "kill" the word (P22, P32). - **Documented examples versus argument.** The essay states a small set of concrete examples: the phrase "he unalived himself" heard aloud (P3), the way people refer to the central crime in a Cornell University case (P4), and a toddler saying "please like and subscribe" (P10). It also states facts about people and companies: the Ellison family and TikTok, and Musk and child sexual abuse material (P12). The reasoning that connects them (a deliberate system, a lost generation, harm to AI training data) is the author's argument, not a documented finding. No study, platform policy, or moderation log is cited in the capture. - **Why it matters.** If the author is right, content-warning tools and keyword filters stop working because the vocabulary keeps shifting (P6). If the claim is a perception, as P2 and P28 allow, the habit could persist with no enforcement behind it. - **Key stakeholders.** Writers, artists and activists who post on algorithmic platforms; younger people the author says grew up inside those platforms; readers who use content filters; platform operators named in the essay (TikTok, Meta, Google, Instagram, YouTube, X, Substack, Spotify); and people who write about or are affected by death, suicide, and sexual assault. - **Risks and tradeoffs.** The author's own instruction carries risk for the people who follow it: P16 quotes the objections "it will cost me audience" and demonetization, and the essay answers with "what if they do." The essay endorses content warnings and abbreviations like "SA" (P5), so the target is euphemism done for the algorithm, not all careful wording. The topic involves suicide and rape, and the essay treats the words as things that should be sayable. Readers who are sensitive to those topics may want to read with that in mind. - **Bottom line.** A forceful commentary with one useful, checkable distinction (warning markers a reader chooses versus wording a platform forces) and a set of unverified causal claims. The author concedes in P2 and P28 that part of the phenomenon may be belief and self-censoring, not enforcement. ## 📖 Main content **The definition and the spread (P1-P3).** Dash opens by saying older people who ask what "unalive" means usually don't know, except those with kids or those who study internet culture (P1). Dash defines it as a word used in place of dying, death, or killing on platforms where saying those words is expected to draw punishment from the algorithm, and adds that this may be perception, since "no one is actually willing to find out" (P2). The author says the pattern started on TikTok, spread to nearly all digital media, and now appears in speech, with the example of hearing someone say "he unalived himself" about a death by suicide (P3). The sentence is reported as the author's account of what Dash has heard. **A second example and the author's limits on the argument (P4-P6).** The author says the pattern became more visible during recent online discussion of a Cornell University rape case, where people write r*pe, "grape," or the grape emoji (P4). Dash then states what they are not against: content warnings, sensitivity markers, abbreviations such as "SA," and ironic distance as a coping tool (P5). Dash's objection is to wording chosen to appease platforms, because a filter user's tool "will no longer be effective" once everyone uses misspellings and imagery (P6). **Thesis and the "constraints" argument (P7-P13).** P7 states the thesis: people spread mangled syntax because they believe or have experienced silencing for plain description. P8 is a subhead. P9 argues that a generation hasn't noticed how extreme it is that platforms chose its language, and that algorithm demands pushed creators to make exaggerated thumbnail faces. P10 uses the parental anecdote of a toddler saying "please like and subscribe." P11 describes Instagram users begging for cultish comment phrases to earn AI-generated replies, with the algorithm described as "an angry god." P12 argues that the people who pretended to defend free expression now accept billionaires deciding allowed words, names Larry Ellison (TikTok) and Elon Musk (X), and links the author's own guide to leaving X. P13 argues that AI tools trained on this internet may not know the uncensored world, with "genocide" as the example. **The creative argument and the answer (P14-P22).** After a second subhead (P14, profane in the source), P15 argues that imagination is limited by the words available, so work is "unalive before it is even born." P16 gives the answer, "say what you mean," and admits it takes courage and possible sacrifice. P17-P19 ask whether readers will push, move platforms, or use the ones they already have (email newsletters, podcasts, the author's own blog), and add the aside that Substack is not an email newsletter and a Spotify show is not a podcast. P20 speaks to readers from the 20th century: MySpace styling, open-source LiveJournal code, and cheap blogs show another world is possible. P21 speaks to readers from this century and calls the situation "a system working as designed." P22 closes the section with "It is time to kill “unalive”." **The six-step response and the close (P23-P32).** P24-P30 list actions: start from what you mean, build a presence on every platform, tell the audience free speech matters more than corporate convenience, build alliances, release uncensored work and see where platforms push back, make noise if a platform limits you, and praise others who push back. P28 adds that readers may find they were censoring themselves against limits that weren't there. P31 warns of rising authoritarianism and muted artists if people keep accepting forced words. P32 closes on the author's anger at tech executives and repeats that it's time for "unalive" to die. **What the essay does and doesn't establish.** It establishes the author's observations (what Dash has heard, what people write) and the argument. It does not establish how any platform treats these words, whether a given word triggers moderation, how many people use euphemisms, or what AI models know about topics that platforms discourage. The captured text offers no platform policy, test, or dataset for those points. ## 🗒️ Notes #### 📄 ¶1-¶3 What "unalive" is and how far it has spread - ¶1: Older readers usually don't know the word, except parents and people who study internet culture. This is the author's observation, with no count. - ¶2: Defines "unalive" as a stand-in for death or killing on platforms where the algorithm is expected to punish the real words, then adds that this may only be perception. - ¶3: Places the origin on TikTok and algorithmic feeds, says the habit now appears in speech, and gives the example of hearing "he unalived himself" about a suicide. #### 📄 ¶4-¶7 A second example, the author's limits, and the thesis - ¶4: Cites online talk about a Cornell rape case, where people write r*pe, "grape," or the grape emoji. The author notes the assumption, not the certainty, that the word can't be said. - ¶5: Endorses content warnings, "SA," and ironic distance. This is the distinction the argument depends on. - ¶6: Argues that wording chosen to appease platforms defeats user filters. - ¶7: States the thesis: mangled syntax spreads because people believe or have experienced silencing for plain language. #### 📄 ¶8-¶13 "You Were Not Born With These Constraints" - ¶9: A generation grew up without noticing that platforms chose its language, and creators were pushed into exaggerated thumbnail faces. - ¶10-¶11: Two scenes: a toddler saying "please like and subscribe," and Instagram users begging for scripted comments to feed an algorithm. - ¶12: Names Ellison and Musk as examples of powerful people who now control speech, and links the author's own guide to leaving X. - ¶13: Argues AI systems trained on a self-censoring internet may not know the uncensored world, using "genocide" as the example. #### 📄 ¶14-¶19 The creative case and "say what you mean" - ¶14-¶15: After a profane subhead, argues imagination is limited by available words, so work is "unalive before it is even born." - ¶16-¶17: The answer is "say what you mean," with the objections (lost audience, lost followers, demonetization) quoted and answered with "what if they do." - ¶18-¶19: Points to places the reader controls: email newsletters, podcasts, the author's blog. The aside says Substack is not an email newsletter and a Spotify show is not a podcast. #### 📄 ¶20-¶22 Two audiences and the call - ¶20: Speaks to readers of the 20th century: MySpace styling, open-source LiveJournal code, and cheap blogs read by millions show another world is possible. The author says that last point is still possible. - ¶21: Speaks to readers of this century: the best artists are invisible, activists self-censor, and "it's not a conspiracy, it's a system working as designed." The author's claim, not a documented finding. - ¶22: The call: "It is time to kill “unalive”." #### 📄 ¶23-¶30 The seven-part response - ¶24-¶25: Start from what you mean, not from what a platform permits, and build a presence on every platform even where you have fewer followers. - ¶26-¶27: Tell the audience free speech matters more than corporate convenience, keep it positive, and build alliances with others who share the values. - ¶28-¶29: Publish uncensored work and see where platforms push back, since self-censoring may anticipate limits that aren't there. If a platform does limit you, make noise: press, alliance, other platforms, and amplify the compliant parts of your work. - ¶30: Find others pushing back and praise their work. #### 📄 ¶31-¶32 Stakes and close - ¶31: Warns that accepting forced words while X allows hateful content leads to authoritarianism and muted artists, and says this breaks Dash's heart. - ¶32: Says Dash gets angry about tech executives who boast about ending the world or call for ethnic cleansing while restricting ordinary speech, and repeats the call. ## 📚 References The link list for this capture has 9 entries (anchor text -> URL), supplied with this task. It carries no paragraph numbers, so the placements below follow list order against the paragraphs where each anchor phrase appears. Eight of the nine links point to earlier posts on the author's own site. The ninth is a Wikipedia article. **Linked sources (from the captured link list)** - "rape case at Cornell University" (P4): https://en.wikipedia.org/wiki/2024_Cornell_University_rape_allegations. This is the only outside source in the list and the one source for the case the essay mentions. The essay says the case involves "horrific lack of accountability," which is the author's framing. This analysis did not read the Wikipedia article and claims nothing about its contents. - "TikTok" (likely P12, by list order): https://www.anildash.com/2025/10/06/the-threat-to-video-creators/ - "generate and distribute child sexual abuse material for profit" (P12): https://www.anildash.com/2026/02/18/threatening-kids-with-ai/. The claim about Musk is the author's, linked to the author's own earlier post. - "a guide" (P12): https://www.anildash.com/2026/02/23/taking-action-ai-harms/. The essay says this guide covers getting an organization off X. - "podcast" (likely P18, by list order): https://www.anildash.com/2026/02/28/apple-video-podcast-power/ - "Substack" (P19): https://www.anildash.com/2024/11/19/dont-call-it-a-substack/ - "podcast" (likely P19, by list order): https://www.anildash.com/2024/02/05/wherever-you-get-podcasts/ - "another world is possible" (P20): https://www.anildash.com/2026/03/27/endgame-open-web/ - "a system working as designed" (P21): https://www.anildash.com/2024/05/28/systems-the-purpose-of-a-system/ **Claims with no link in the captured markup.** The claims that platforms punish or suppress the plain words (P2, P7), that "unalive" originated on TikTok and spread to speech (P3), that filter tools are defeated by euphemism (P6), that creators were forced into thumbnail faces by algorithms (P9), that Larry Ellison's family controls speech on TikTok (P12), that AI tools trained on this internet lack knowledge of uncensored topics (P13), and that executives called for ethnic cleansing or boasted about ending the world (P32) have no link in the captured markup. The essay presents the first two as observation and belief, and the rest as the author's assertion. The captured text shows no platform policy, study, or quote for any of them. **Self-citation pattern.** Eight of nine links are to the author's own earlier posts, so the links show the author's prior views, not independent confirmation. That is normal for a personal blog, and it means a reader can follow the argument's history but can't verify the claims from the links alone. ## 🗣️ Paraphrase Anil Dash writes that "unalive" and its relatives (r*pe, "grape," the grape emoji) exist because people think platforms will punish plain words for death and assault. Dash says the habit began on TikTok, spread to most digital media, and now appears in speech. Dash agrees content warnings and abbreviations like "SA" are fine, but says most substitutions are made to appease an algorithm, and they break filters for readers who want to avoid the topics. The author argues that a generation grew up inside platform rules without noticing them, that powerful owners benefit from controlling speech, and that AI tools trained on this internet may not know the uncensored world. Dash's answer is to say what you mean, keep other places to publish (email, a blog, a podcast), publish uncensored work to see where platforms push back, make noise if they do, and support others doing the same. The essay admits that part of the problem may be self-censoring against limits that don't exist, and that the plan costs reach. ## 📰 Newsletter **A case against writing "unalive."** Anil Dash's essay of October 6 argues that euphemisms like "unalive," r*pe, and the grape emoji spread because people assume platforms punish plain words, and that the habit is now leaking into speech. Dash draws a line: content warnings and tags such as "SA" are fine because they give readers control, while wording chosen to appease an algorithm hurts readers who rely on filters, since misspellings defeat them. The essay is opinion, in strong language, and mixes a few concrete examples (a phrase heard aloud, a toddler saying "please like and subscribe") with large claims about platform owners and AI training data that it doesn't source in the captured text. Its practical advice is to say what you mean, publish on channels you control, test where platforms push back, and make noise if they do. It admits some self-censoring may anticipate limits that aren't there. Read the original: [anildash.com](https://www.anildash.com/2026/10/06/kill-unalive/). This summary doesn't give mental-health guidance and takes no position on the topics' wording beyond what the author argues. ## ✅ Action items These items concern the reader's own writing, documentation, and community practices. They apply the essay's argument to project work and are not mental-health guidance. #### DevRel - Developer advocacy: when a talk or post covers sensitive topics, use a stated content note that a reader can choose to skip, in the way P5 describes, and avoid misspelled substitutes that defeat search and filters. - Education: add a short note to writing guidelines on the difference between a content warning (reader's control) and a platform-driven euphemism, and show one example of each. - Developer experience: no direct action from this source; nearest relevance is keyword-searchable docs, where "unalive"-style substitutions would break search. - Developer marketing: for posts on algorithmic platforms, keep the canonical version on a site you control (P18) and link out. - Community management: when moderating, decide whether to enforce euphemism or plain wording and state it in the rules, with a reason. - Internal/external: if a platform limits a post, log what happened and the platform's stated reason before deciding whether to speak publicly (P29). #### OSPO - Policy and governance: review whether community moderation rules ask members to rewrite plain words into euphemisms, and if so, why. - License compliance: no direct action from this source. - Security management: no direct action from this source. - Software inventory: no direct action from this source; nearest relevance is keeping a list of the channels the project owns versus rents (P18-P19). - Developer support: keep a page of channels the project controls (site, mailing list, feed) so announcements don't live only on platforms. - Community engagement: no direct action from this source; nearest relevance is the "build alliances" step (P27). - Strategic alignment: weigh the cost the essay names (audience, demonetization) against owning distribution. #### Community - Maintainers: write a one-paragraph content-note convention for issue threads that discuss harm, abuse, or safety topics. - Contributors: no direct action from this source. - Sponsors: no direct action from this source. - Ecosystem: no direct action from this source; nearest relevance is the essay's point about owned channels (P18). #### WordPress - Core: no direct action from this source. - Core Performance: no direct action from this source. - Core AI: no direct action from this source; nearest relevance is the training-data claim in P13, which is the author's assertion. - Core Program Team: no direct action from this source. - Mobile: none. - Playground: none. - Community: none directly; see the Community headings above for a content-note convention. - Training: if a lesson covers writing about sensitive topics, teach content notes versus euphemism, with sources. - Marketing: none directly; the essay's owned-channel point (P18, P20) applies to site-owner messaging. - Support: no direct action from this source. - Design: none directly; content-note patterns (collapsible, labeled) are a design item. - Documentation: when documentation mentions harmful content or safety topics, use plain words with a content note where appropriate. - Photos: none. - TV: none. - Polyglots: none from this source; euphemisms vary by language and the essay doesn't address that. - Accessibility: none from this source. - Meta: none. - CLI: none. - Hosting: none. - Tide: none. - Openverse: none. - Themes: none. - Plugins: none. #### The WPOCC - Programs: no direct action from this source. - Projects: no direct action from this source. - Funding experiments: none. ### FAIR - Technical independence: no direct action from this source; nearest relevance is the essay's argument for channels the publisher controls (P18, P20), which parallels decentralized distribution. - Policies and covenants: no direct action from this source. - Productized governance artifacts: no direct action from this source. ## 🧪 Hypotheses 1. A plain-word post and a euphemism post on the same platform, with the same account and subject, show no measurable difference in reach on a majority of tests, which would support P28's point that self-censoring anticipates limits that aren't there. 2. Content filters that match plain keywords miss a large share of posts that use euphemisms or emoji for the same topics, supporting P6. 3. Writers who move their primary publishing to a self-controlled channel keep a larger share of their audience than writers expect before moving. 4. People who use "unalive" in speech use it more with peers who are heavy users of algorithmic-feed platforms than with peers who aren't (testing P3). 5. Language models trained on text with substitution spellings handle direct questions about the substituted topics worse than models trained on plain text (testing P13). 6. Readers shown the post's content-warning distinction (P5) correctly separate "warning marker" from "platform euphemism" in new examples. ## 📊 Methodology - **H1.** Create paired accounts or paired scheduled posts on one platform with matched content, one using plain words and one using substitutes, and compare reach, flags, and removals over a fixed period. Respect each platform's terms, avoid content that could cause harm, and record results as observations, not policy claims. - **H2.** Assemble a test set of posts on a topic using plain terms and common substitutes, run them through two or three open keyword filters, and report recall for each form. - **H3.** Survey writers or creators before and after a move to an owned channel, asking for follower counts, engagement, and revenue on each channel. Report self-reported numbers separately from platform analytics. - **H4.** Survey a sample of speakers by age and platform use on how often they say substituted words aloud, and compare groups. Use anonymous responses and avoid asking about personal experiences with the topics. - **H5.** Prompt models with matched plain and substituted questions on the same facts, score answer accuracy against a reference, and compare. Note that model training data composition is usually not public. - **H6.** Show readers 20 paired examples (a content note, a euphemism) and ask them to label each. Score agreement with the author's distinction. ## ❓ Questions - Which specific platforms have stated policies that penalize plain words like "dead," "kill," or "rape," and where are those policies written? - How much of the substitution habit is enforcement, how much is belief, and how much is peer norm (the essay allows all three)? - What does the essay propose for readers who rely on content filters, beyond asking writers to stop? - How does the call to say plain words interact with writers and audiences who are affected by suicide or sexual violence and who prefer softer terms for their own reasons (the essay allows some in P5)? - Does the "constraint" argument in P9 apply to older users on algorithmic platforms, or only to the generation the author describes? - What evidence supports the claim that major AI tools lack knowledge of topics that platforms discourage (P13)? - Which of the essay's "response" steps are low cost, and which require losing audience or revenue? - What would count as a refutation: a platform audit showing plain words aren't penalized, or a user study showing the habit is mostly a norm? ## 🔮 Future research - **Enforcement versus belief.** Study how much euphemism use follows documented enforcement and how much follows folk belief, using platform transparency reports where they exist. - **Filter effectiveness.** Measure how keyword and classifier filters perform on substituted spellings and emoji over time. - **Language drift.** Track the spread of "unalive" and similar terms across platforms and into speech, with dates and age groups. - **Training data effects.** Test whether substitution-heavy training text affects model answers on the topics involved. - **Owned-channel outcomes.** Collect case studies of creators who moved or diversified channels and what happened to reach and income. ## 💡 Implications - **Practice.** Separate "content note" from "euphemism." The first gives readers a choice. The second may break their tools if it changes the words, per the author's argument (P5-P6). - **Policy.** Platform and community moderation rules that are unclear push people to guess. A published, specific list of what is and isn't penalized would test the essay's perception claim (P2). - **Security.** Not a security topic. The one tangent is account and channel ownership: relying on one platform for distribution is a single point of failure (P19). - **Education.** The essay gives a usable case for teaching how word choice, platform incentives, and audience tools interact. It doesn't offer guidance on how to discuss suicide or sexual violence, and this analysis doesn't either. - **Funding.** The author names the cost of the plan: lost audience, lost followers, demonetization (P16). The essay gives no figures. ## 🧠 Learning styles assessment - **Reading level estimate.** Flesch-Kincaid Grade Level of about 10 and Flesch Reading Ease of about 58, from a simple syllable heuristic on the 1,926-word text. Sentences run long, with parenthetical asides, and the vocabulary is plain. Profanity and rhetorical questions keep the register conversational. #### 👀 Visual learners - Present: the capture has no figure. The 🍇 emoji (P4) is the only image-like element in the text. - Improve: a small table of "warning marker" versus "platform euphemism" examples would make the P5-P6 distinction visible. #### 🔊 Aural learners - Present: the essay's voice is spoken-style, with direct address and short punchy lines (P17, P22), which reads aloud well. No audio is referenced in the captured text. - Improve: an audio version would help listeners and could demonstrate the point about plain speech. #### 🗣️ Verbal learners - Present: short paragraphs, direct address, and two subheads (P8, P14) support linear reading. - Improve: define "algorithmic feed" and "demonetize" in one clause each for readers unfamiliar with them. #### 🧩 Kinesthetic learners - Present: P23-P30 give a seven-step action list. - Improve: add one concrete exercise, such as rewriting a euphemistic post in plain words and posting it where the writer can observe the result. #### 🧠 Logical learners - Present: a clear thesis (P7), a conceded limit (P5), and an explicit conditional about perception (P2, P28). - Improve: separate observation, belief, and assertion with labels, and cite a source for each factual claim about platform behavior. #### 🧑‍🤝‍🧑 Social learners - Present: the call to build alliances, praise others, and make noise together (P27-P30) is a social strategy. - Improve: no discussion prompt or community link appears in the captured text. #### 🧘‍♀️ Solitary learners - Present: self-contained, with the definition of "unalive" in P2 for new readers. - Improve: a short summary box at the top would help readers who skim. #### 🌿 Naturalistic learners - Present: none; the content concerns language, platforms, and politics. The "angry god" metaphor (P11) is the only nature-adjacent image and is figurative. - Improve: not applicable beyond that. ## 📊 Accessibility assessment Judged from the captured text and link list only. Rendered contrast, focus order, and image alt text are unknowable from this capture. #### Screen reader semantics - Present: two short standalone lines (P8, P14) read as subheads, which suggests a heading structure. The captured text doesn't show heading levels. - Improve: confirm they are real heading elements (WCAG 2.2 AA 1.3.1 Info and Relationships, 2.4.6 Headings and Labels). The 🍇 emoji in P4 and the censored spelling r*pe are read aloud differently by different screen readers, and the asterisk may be announced as "star" (1.3.1, 3.1.5 Reading Level is AAA and not required). The essay's subject, euphemism, makes this a topical check. #### Keyboard navigation - Present: 9 links in the list, all ordinary anchors as far as the capture shows. - Improve: not verifiable from text. Check 2.1.1 Keyboard and 2.4.7 Focus Visible on the rendered page. #### Contrast and readability - Present: unknowable from the capture. Paragraphs are short to medium. - Improve: verify 1.4.3 Contrast (Minimum) on body text and link color. #### Text resizing and reflow - Present: text-only content should reflow, and the capture contains no fixed-size media. - Improve: verify 1.4.4 Resize Text and 1.4.10 Reflow at 320 CSS px and 200% zoom. #### Media and captions - No media referenced in the captured text. The text doesn't mention captions or transcripts. The emoji in P4 carries meaning in context (it stands for a word), so a text equivalent is present in the same sentence (r*pe and "grape"). #### Forms and interactions - Present: no form appears in the captured text. - Improve: if the page has comments or a signup, check 3.3.2 Labels or Instructions. #### Standards reference - Likely gaps to verify against WCAG 2.2 AA: **2.4.4** Link Purpose (In Context), because anchors like "a guide," "podcast" (twice, pointing to two different posts), and "TikTok" are short, and the two "podcast" links share anchor text but go to different URLs. That can fail 2.4.4 or 3.2.4 Consistent Identification if context doesn't tell them apart. **1.3.1** for the apparent headings. Everything else needs the rendered page. ## 🌍 DEIB assessment #### Cultural breadth - Present: the essay is written from a US, English-language, platform-heavy frame (TikTok, Instagram, Cornell, MySpace, LiveJournal). - Improve: the text doesn't address languages other than English or regions with different platform rules. A note on that would widen relevance. This is a gap in context, not an error. #### Inclusive language - Present: the essay endorses content warnings and "SA" (P5), so it recognizes reader choice on disturbing topics. It also grants that ironic distance helps some people cope with painful subjects (P5). - Improve: the essay uses profanity, insults a group of owners and executives, and calls older and younger readers "relics" and people "fucked over" (P20-P21). Readers who find strong language a barrier may stop reading. The wording on suicide ("he unalived himself") is quoted as an example of the habit, with no stated view on how a person affected should talk about it. The analysis notes this and gives no guidance. #### Neurodiversity support - Present: short paragraphs and clear action list aid scanning. - Improve: the register shifts between anecdote, rant, and instruction. A short summary and the action list near the top would help readers who lose the thread. #### Representation - Present: the essay names Larry Ellison and Elon Musk, uses pronouns for them where the source does ("his family," "he wants"), and speaks to "old people" and "this century" in generational terms. It doesn't name individual affected people. The author uses first person. - Improve: no voices from creators who rely on euphemisms, moderators, or people affected by the topics appear. That is a typical limit of a single-author commentary. #### Psychological safety - Present: the essay gives content-warning endorsement (P5) and opens with a plain definition. - Improve: it contains a direct reference to suicide (P3) and to a rape case (P4) and uses strong language and anger throughout. Readers who find these topics hard may want to know before reading. This analysis gives no mental-health guidance and adds none. ## ✅ Logical validity This is opinion and advocacy. The essay argues a thesis and a plan, so fallacy checks apply to its reasoning, with one caveat: it flags uncertainty about its own premise in places (P2, P28). #### Strong reasoning - **A clear distinction, ¶5-¶6.** The author separates content warnings (reader control) from platform-driven euphemism, and ties the harm to a specific mechanism: misspellings defeat filter tools. A reader can test this by running a filter on both forms. - **A conceded premise, ¶2 and ¶28.** The author says the censorship may be perceived and that people may be censoring themselves in anticipation of limits that aren't there. This keeps the essay from resting wholly on an unproven enforcement claim. - **A testable instruction, ¶28.** "Start releasing your work uncensored and see where the platforms push back" is an experiment that can falsify the author's own belief. - **Named costs, ¶16.** The objections ("it will cost me audience," demonetization) are quoted and addressed instead of ignored, though the answer is "what if they do." - **Owned-channel logic, ¶18-¶20.** The argument that email, a blog, and a podcast are channels no algorithm controls is a mechanism, and the author's earlier posts (linked) back the distinction between email newsletters and platform products. #### Fallacies - **Overgeneralization, ¶9 and ¶10.** "An entire generation" and "every parent of my cohort" generalize from anecdote (a toddler saying "please like and subscribe") to groups. *Fix:* "many parents I know," with a source if a number exists. - **Motive attribution, ¶9.** "People who hate that generation’s ability to express itself" assigns hatred to platform operators with no evidence in the capture. *Fix:* describe the incentive (engagement targets) without asserting hatred, or cite statements. - **Unsupported absolute, ¶13.** "If you can’t say 'genocide' on any of the major platforms" is stated as fact, then used to infer that AI tools "aren't very likely to know much about genocide." P2 itself allows the restriction may be perception only, so the inference is weaker than the sentence. *Fix:* cite a platform policy and a test of a model's answers. - **Guilt by association, ¶12.** The paragraph links people who contrived campus free-speech examples, "cancel culture," and billionaire speech control as one group. The essay doesn't say who is in that group beyond two named executives. *Fix:* name the specific actors and their statements. - **Slippery slope, ¶31.** Accepting platform wording leads to authoritarianism and "harms against the vulnerable accelerate." The chain isn't shown. *Fix:* state the steps or soften to a risk. - **Dismissal of the cost, ¶17.** "Okay, what if they do?" treats lost income as an answer to the objection. For a creator who relies on that income, it isn't. The author does admit "some sacrifice" (P16). *Fix:* acknowledge who can afford the risk and offer lower-cost steps first, as P25-P26 partly do. - **Absolutes without sources, ¶21 and ¶32.** "It’s not a conspiracy, it’s a system working as designed" and the claims about executives' public statements have no link in the captured markup. The "system working as designed" anchor links to the author's post on the topic, which this analysis did not read. Where an item above is the author's opinion, it is labeled as the author's argument in the sections before this one. #### Tone of voice - Angry, direct, and persuasive, with profanity, second-person address, and short punchy lines (P17, P22). The tone is advocacy, not neutral analysis, and the author says they get angry (P32). It softens in P5, where Dash grants content warnings and ironic distance. Outlet viewpoint: a personal blog with the author as publisher, no sponsor shown in the capture, and the author's own commercial interest isn't stated. Dash recommends blogs and email newsletters, which are the kind of site this is. ## 🛡️ DARVO analysis Not applicable to this source. DARVO (Deny, Attack, Reverse Victim and Offender) describes a response pattern to an accusation. This is an advocacy essay, not a reply to one. Each prong is checked for the author's voice anyway. - **Deny.** The essay doesn't respond to a charge against the author. P2 and P28 concede that the censorship may be perceived or self-imposed, which is a limit on the author's own claim, not a denial. - **Attack.** The essay attacks platform owners and executives in strong terms (P9, P12, P32) and calls readers' habits "bullshit" and "fucked over" (P21). This is criticism of named public actors and of a system, in a commentary, not a reply to an accusation. - **Reverse victim and offender.** The essay frames platform users, especially young ones, as the wronged party and platform owners as the offender (P9, P12, P21). That is the thesis, not a reversal of an accusation. It is the author's argument and isn't supported by documents in the capture. ## 🎯 Conclusion Anil Dash's essay argues that "unalive," r*pe, and the grape emoji spread because people assume platforms punish plain words, and that this harms people who use filters and trains a generation to let platforms pick its words. Dash asks readers to say what they mean, keep channels they control, test where platforms push back, and make noise when they do. The most useful part is the distinction between content warnings that give a reader control (P5) and wording changed for an algorithm (P6). The most important limit is the author's own: P2 and P28 allow that the restriction may be perception or self-censoring. What the essay doesn't establish: how any platform actually treats these words, how widespread the habit is, whether AI tools lack knowledge on the topics (P13), or the intent attributed to owners (P9). Eight of its nine links go to the author's own earlier posts, and none go to a platform policy or study. The essay touches suicide and sexual assault. This analysis reports the author's argument, adds no guidance on those topics, and does not give mental-health guidance. **Next steps for readers:** find the written policy of the platform you use before assuming a word is penalized, and test one plain-word post if the cost is acceptable to you. Keep content warnings, which the author endorses, and treat the author's claims about owners and AI as the author's argument until sourced. ## 🔄 Refactor Appendix — We are going to kill “unalive” > Density: 2–5 Literature atoms (H3 pad), 0–3 Permanent candidates (H4 pad). > Emitted: 5 Literature, 2 Permanent. Resonance filter applied: the Larry Ellison, > Elon Musk, and AI-training-data claims were dropped as unsourced assertions. > > Konik claim-header rules: present tense, declarative, ≤10 words, searchable, unique. > > Atomicity: one declarative sentence per atom. "and"/"also" → split. > > Tag pool: #bmf · #addendum · #xref · #articleseed · #storystem · #share · #aphorisms ### 📚 Literature Split Pad #### [H1] The author calls for ending "unalive" > [!quote] [¶22] > It is time to kill “unalive”. **Context & use**: - ¶22, the thesis line, repeated in different words at P32. The essay's title echoes it. - A call to action, not a finding. Use it as the statement of the author's position. **Tags**: #xref #articleseed  #### [H2] The restriction may be perception only > [!quote] [¶2] > Or, maybe, where the perception is that using those words will result in being censored by the algorithm **Context & use**: - ¶2, inside the definition. The author separates enforcement from belief. - Keep this next to H4. Together they show the author's own limit on the enforcement claim. **Tags**: #xref  #### [H3] Euphemism defeats content filters > [!quote] [¶6] > If someone is using a tool to filter out content, it will no longer be effective **Context & use**: - ¶6, the mechanism for harm to readers who use filters. The author's claim, with no test in the capture. - Checkable: run a filter on plain and substituted spellings. **Tags**: #xref #share  #### [H4] Self-censoring may anticipate absent limits > [!quote] [¶28] > sometimes you were censoring yourself in anticipation of limits that weren’t even there. **Context & use**: - ¶28, in the author's "release uncensored work" step, as a reason to test the limit. - Pairs with H2. It makes the author's plan an experiment. **Tags**: #xref #aphorisms  #### [H5] The remedy is to say what you mean > [!quote] [¶16] > The answer to this is simple: say what you mean. **Context & use**: - ¶16, the core instruction. The author admits it takes courage and possible sacrifice in the next sentences. - Use with H3: Dash's own distinction in P5 keeps content warnings. **Tags**: #xref #share  ### 🧠 Atomic Split Pad #### [P1] A warning marker and a platform euphemism are different things **Claim** (one sentence, Courtney's voice): A content warning gives the reader a choice, and a euphemism written for an algorithm removes the choice by changing the words a filter looks for. **Evidence pointer**: [[#H3]] · [[#H5]] **Scope & conditions**: Holds when readers use keyword filters. The author endorses warnings and "SA" in P5. Weakens where a euphemism is a personal coping choice, which the author also allows. Treat the filter-defeat claim as the author's until tested. **Links**: - [[Content notes and warnings]] — because extends - [[Keyword filters and substitution spellings]] — because evidence-for the mechanism - [[Platform moderation folk beliefs]] — because contrasts on enforcement versus perception  #### [P2] Test the platform limit before you obey it **Claim** (one sentence, Courtney's voice): Before rewriting a post to dodge an algorithm, publish the plain version once and see whether the platform responds, because the limit may be a belief and not a rule. **Evidence pointer**: [[#H2]] · [[#H4]] **Scope & conditions**: Holds when the cost of a limited post is low. Weakens for creators whose income depends on reach, a cost the author admits in P16. Doesn't apply to topics where a platform's written rules are known. **Links**: - [[Owned channels versus rented platforms]] — because applies-in - [[Self-censorship under uncertainty]] — because extends - [[Platform policy transparency]] — because contrasts on guessing versus reading the rule  ### ✏️ Fleeting Queue - [F1] Find each major platform's written policy on terms for death, suicide, and sexual assault and note what they actually say — https://www.anildash.com/2026/10/06/kill-unalive/ #fleeting #inbox - [F2] Run a keyword filter on plain and substituted spellings and compare recall #fleeting #inbox - [F3] Open question: how much substitution is enforcement, how much belief, and how much peer norm? #fleeting #inbox - [F4] Article seed: "Content note versus euphemism," a short explainer using P5-P6 as the example #fleeting #inbox #articleseed - [F5] Check the author's linked post on systems and design (P21) before citing the "system working as designed" line #fleeting #inbox ### → Things 3 (max 2 actions) - [ ] Look up the written moderation policies of the platforms I use and list what they say about these terms — research, 30 min #things - [ ] Draft a one-paragraph content-note convention for my own posts and docs — writing, 20 min #things  📄 **Full text:** [[Full Document Contents/Articles/We Are Going to Kill “Unalive”.md|We Are Going to Kill “Unalive” by Anil Dash]] ## Highlights - This will, of course, probably seem very weird to most people who were born in or grew up in the 21st century. Just to recap for the olds: “unalive” is the word you use to represent concepts like dying, or death, or killing or being killed, on digital platforms where saying those words accurately will cause the algorithm to punish or censor you. Or, maybe, where the perception is that using those words will result in being censored by the algorithm, and no one is actually willing to find out what happens if you use the forbidden words. ([View Highlight](https://read.readwise.io/read/01m4b9ww9f7ep5hv9rd8w2dhtw)) - Note: [H2] The restriction may be perception only - Tags: atom/literature/h2 - But most of the different variations of wording and emojis are coming from trying to appease the platforms, and there’s a heavy cost for those who are worried about being mindful: If someone is using a tool to filter out content, it will no longer be effective because everyone is using misspellings and euphemisms and imagery to get around the algorithm. ([View Highlight](https://read.readwise.io/read/01m4b9wymrmgpksfmxwsy5qm3b)) - Note: [H3] Euphemism defeats content filters - Tags: atom/literature/h3 - The answer to this is simple: say what you mean. This will take, to some degree, courage. It may even take, I hesitate to say, some sacrifice. When I suggest this course of action to people, they inevitably say, “But it will cost me audience!” or “But what if I lose followers!” or “What if they demonetize me!” ([View Highlight](https://read.readwise.io/read/01m4b9x3d1sfkk1me1hz3k6p4p)) - Note: [H5] The remedy is to say what you mean - Tags: atom/literature/h3 - It should make you furious. It is time to kill “unalive”. ([View Highlight](https://read.readwise.io/read/01m4b9wsy88gwp17mnp0s34dax)) - Note: [H1] The author calls for ending "unalive" - Tags: atom/literature/h1 - - Start releasing your work uncensored and see where the platforms push back. (You may be surprised: sometimes you were censoring yourself in anticipation of limits that weren’t even there.) ([View Highlight](https://read.readwise.io/read/01m4b9x15d9vg4qknep7qj5eyd)) - Note: [H4] Self-censoring may anticipate absent limits - Tags: atom/literature/h3