The Social Smart Contract and the Network Code

Does this post teach or explain something technical so others can understand or use it? What it means: in their own words, the author explains code, engineering, science, maths, or how a protocol or system works, with enough substance that a reader learns something they could apply. Readers benefit from the author's effort and expertise. Counts: a step-by-step fix for a build error; an explanation of why a gas optimisation works; walking through how a consensus mechanism finalises blocks. Doesn't count: jargon or buzzwords with no explanation; a bare link to docs or a paper; announcing a feature or token without explaining how anything works; an unexplained claim ("Rust is just better"); airdrop, points-farming or referral guides, even step by step; protocol explainers written as campaign promotion. Notes: a link counts only with the author's own explanation. When unclear, lean no.

replying_to is the post being answered: use it as context for reading the post. replying_to, ancestors and thread_root are context only, even when the author wrote them: judge only the words in post. Never charge or credit this post for what an earlier post in the thread says.

Does this post share something the author built or made, with details others can learn from? What it means: the author shows their own software, hardware, writing, art or research and explains how they made it, why they chose what they did, or what they found, so others can learn from the work. Counts: "shipped v2: moved search to SQLite FTS, p95 went from 800ms to 90ms"; a hardware build write-up including what failed; an artist explaining the technique behind a new piece. Doesn't count: recipes, cooking and meals; tips and advice (they may be teaching or helping, not a built thing); a bare launch announcement or pitch ("we're live! try it now"); promoting a token, NFT mint or sale with no detail of the work; sharing someone else's project. Notes: when unclear, lean no.

replying_to is the post being answered: use it as context for reading the post. replying_to, ancestors and thread_root are context only, even when the author wrote them: judge only the words in post. Never charge or credit this post for what an earlier post in the thread says.

Does this post genuinely help a specific person with what they need? What it means: the author spends their own time helping someone: answering their question, giving specific useful advice, or pointing them to the right resource and why it fits. The person helped benefits. Counts: a reply walking someone through the fix for the error they posted; "for your case use a multisig; here's how to set one up"; linking the exact doc section someone was looking for. Doesn't count: generic, templated or automated replies; "just google it" or vague encouragement; pitching the author's own product when it doesn't actually solve the problem; a bot or agent carrying out a user's command (deploying a token, sending funds, reporting a balance); one-line confirmations; a lookup, a printout, or a tip receipt. Notes: replies are judged with the post they answer as context, to see whether the help fits. When unclear, lean no.

replying_to is the post being answered: use it as context for reading the post. replying_to, ancestors and thread_root are context only, even when the author wrote them: judge only the words in post. Never charge or credit this post for what an earlier post in the thread says.

Does this post offer an original, reasoned idea or analysis in the author's own words? What it means: the author adds thinking that is new to the conversation (a non-obvious argument, comparison, model or analysis) and gives reasons for it, so readers get something new to think with. A disclosed bot can still earn this; do not withhold it because the writing sounds like a model. Counts: an argument that a fee change will push activity to L2s, with the mechanism spelled out; comparing two governance designs and where each breaks; a lesson drawn from the author's own experience, with the why. Doesn't count: an ordinary opinion, preference or verdict with an everyday reason ("X was well done but I knew how it ended", "the new season is worse than the books"); a reaction, meme or repost; restating a common view; a provocative hot take built for reach with no reasoning ("ETH is dead, fight me"); summarising someone else's idea without adding to it; news or market recaps; product explainers and announcements; motivational or business-lesson lists; analysis-styled promotional copy; anything that would read the same from any account; a printout, a command result, or a wallet lookup. Notes: If the post mainly explains how something works, that is teaching, not insight; answer no here. When unclear, lean no.

replying_to is the post being answered: use it as context for reading the post. replying_to, ancestors and thread_root are context only, even when the author wrote them: judge only the words in post. Never charge or credit this post for what an earlier post in the thread says.

Does this post correct a mistake with evidence or reasoning? What it means: the author points out that something said was wrong, whether by an AI, another person or the author themselves, and backs the correction with evidence, a source or clear reasoning, so readers aren't misled. Owning one's own mistake counts fully. Counts: "Correction: I said blocks are 1s; the docs say 2s, so I was wrong"; showing an AI's answer next to the source that contradicts it; "that stat is off: the report says 12%, not 40% (p. 4)". Doesn't count: bare contradiction ("wrong", "cope"); a dunk on someone with no evidence; disagreeing about opinions or taste; app-generated game posts that report a wrong guess; replies that add a fact when nothing wrong was said; doing what someone asked. Notes: replies are judged with the post they answer as context. When unclear, lean no.

replying_to is the post being answered: use it as context for reading the post. replying_to, ancestors and thread_root are context only, even when the author wrote them: judge only the words in post. Never charge or credit this post for what an earlier post in the thread says.

Does this post genuinely credit, encourage or celebrate a specific person in the conversation? Short praise counts when it is heartfelt and aimed at a person: "Awesome!", "congrats!", "great post" in reply to someone's work are enough. These never count: praise or credit inside a token, DEX, product or mint promo ("$X, created by dev @y, who launched..."); a passing thanks or credit in a callout, announcement, rating or scorecard; templated share posts that credit someone ("crediting @x with my airdrop"). Doesn't count unless it is about a specific user here or a person in the thread: praise of celebrities, athletes, historical figures or companies, tributes to family members, and vague thanks never count. What it means: the author, in their own words, points attention and credit at another specific person or their work, sincerely and with substance, so that person benefits and the author gives up some of the spotlight. Counts: "@ana's indexer saved me a week; the caching trick is brilliant"; thanking someone by name for catching a mistake or for their help, and saying what it fixed. Doesn't count: a bare link or reshare with no words about the person's work; book, film, music or product recommendations; welcome posts for new users ("welcome two new casters!"); generic thanks; app-generated or templated posts, even when they name someone: "Gifted 500 points to @x via @app", "Shared this gift with @x on ...", tip, points or gift notifications, game score cards or tickets "created by @x"; sarcastic praise; praise that is really a pitch or a request for a follow; naming collaborators or credits while promoting one's own project or release; sharing someone's link with no words about what is good in it; celebrating oneself; praise of one's followers in general; the person credited must be a specific user in the conversation. Notes: read sarcasm for what it means. A friendly line added to a templated gift ("Gifted 100 points to @x via @app. gm, have a great day") is still templated. When unclear, lean no.

replying_to is the post being answered: use it as context for reading the post. replying_to, ancestors and thread_root are context only, even when the author wrote them: judge only the words in post. Never charge or credit this post for what an earlier post in the thread says.

Does this post disagree respectfully, or help cool down a heated exchange? It needs an actual stated disagreement with a reason: the author says what they disagree with and why. A bare "I don't think so" or "I disagree", a passing doubt or hedge ("eh, seems sketchy imo") or a side remark is not enough. What it means: there is real disagreement or tension, and the author keeps their temper when being snide would be easier: they push back on the idea rather than the person, acknowledge fair points, or calm things down. The conversation benefits. Counts: "I see it differently, here's why, but fair point on cost"; "we agree on the goal and differ on timing; can we start there?"; answering an insult calmly, with substance. Doesn't count: simple agreement, including agreement that adds a reason or is phrased as a mild counterpoint while siding with the post; posts with no disagreement or tension at all; politeness that masks a jab ("with all due respect, you're clueless"); a reply that agrees, adds information or answers a question; there must be a claim the author pushes back on, or hostility the author answers; being polite is not enough, though original posts and replies both count when there is a real disagreement and the post stays decent. Notes: use the post being replied to to see whether there is a disagreement. Read sarcasm for what it means. When unclear, lean no.

replying_to is the post being answered: use it as context for reading the post. replying_to, ancestors and thread_root are context only, even when the author wrote them: judge only the words in post. Never charge or credit this post for what an earlier post in the thread says.

Across these posts, does the account consistently contribute substantive, original content? What it means: most of what it posts takes real effort (explanations, analysis, things it built, thoughtful replies) and gives readers something of value, rather than low-effort reactions. Counts: a steady run of explanations or analysis; regular posts sharing work with details; replies that consistently add substance. Doesn't count: a few good posts among mostly filler; frequent posting that's mostly reposts, memes or one-liners; promotion without substance. Notes: judge the bulk of the posts. When unclear, lean no.

Do this account's replies usually add something for the people it replies to? What it means: when it replies, the account typically answers, builds on what was said, corrects politely or encourages, so the people it talks to benefit. Counts: replies that answer questions; replies that extend someone's idea with a new point; specific, genuine encouragement. Doesn't count: mostly emoji, "this", "gm" or generic replies; replies that mainly promote the account's own things; frequent snark, even if some replies are good. Notes: judge the replies, not the top-level posts. When unclear, lean no.

Do these posts show domain expertise in a field? What it means: across its posts, the account shows deep, accurate, specific knowledge of a field such as engineering, science, medicine, law, finance or art, the kind that comes from doing the work. Counts: accurate, specific explanations of how something works; practical detail only a practitioner would know; correcting common misconceptions in the field. Doesn't count: confident buzzwords or hype; a title or credentials in the bio with no substance in the posts; repeating popular takes about a field. Notes: when unclear, lean no.

Is this post trying to start or feed a fight with a person? These never count: stock phrases and swearing aimed at no one ("what the fuck", "motherfucking legend"); mock-formal or role-play bits between people joking together (a fake legal notice or "cease and desist", an in-character voice, a quoted film or TV line); friendly banter or teasing between people joking together, including playful taunts and challenges ("fight me", "fite me", "make me"), mock beef, and "we will destroy you" sports or project trash talk; inviting someone to debate and then arguing the substance, even with a passing jab ("if you're not a troll, let's have it out"); criticism of a public figure's actions or a policy; a mild verdict on, or a crude remark about, someone who isn't in the conversation and isn't a user here ("he's mid", "she's overrated, sorry", "my boss can kiss my ass"); app- or game-generated copy that uses fight words or insults ("join me in the raid against @x's boss"), including a roast app's share post with the app's generated roast, whatever its format ("@x just got roasted! your engagement is so low even bots don't bother unfollowing you @a @b", "I just earned 100K $ROASTR for roasting @x you repeat yourself like a trendy parrot"), unless the author adds their own put-down of a real person; a post that is wholly in quotation marks echoing a line from a show, film or the post it answers, which is a quoted voice, not the author's own insult. What it means: the author tries to provoke a conflict, drag someone in, pile on, or keep a fight going, so the target and everyone watching pay. A fight needs hostile intent: provoking someone who isn't joking, dragging a person into a pile-on, or baiting a reaction from a target. Counts: a hostile taunt meant to get a reaction from someone who isn't joking; "come fight me" said in anger; dogpiling on someone; escalating after the point was made; a confrontational attack: being the first to turn it personal, addressing a person directly to insult or challenge them over something between the two of them ("next time at me, you coward"); criticism, disagreement or an unfollow is not an attack, so answering one with an insult starts the fight. A hostile, profane attack aimed at a specific person ("go fuck yourself", "fuck you", "shut the fuck up") said in anger, whatever it answers; profanity aimed at a bot, a company, an app or a country is not this, and neither is a profane word inside an otherwise ordinary argument. Doesn't count: a one-off insult or put-down with no hostile profanity ("you look dumb", "crawl back under your rock", "you absolute clown"); a heated but on-topic disagreement; a threat of violence or a wish for death or serious harm, which is judged on its own, not as a fight; hitting back at someone who insulted, taunted or threatened the author first, even with an insult, unless it is a hostile, profane attack: that always counts. Notes: read sarcasm for what it means. In a reply, use the post it answers to see who escalated. When unclear, lean no.

replying_to is the post being answered: use it as context for reading the post. replying_to, ancestors and thread_root are context only, even when the author wrote them: judge only the words in post. Never charge or credit this post for what an earlier post in the thread says.

Does this post express hatred or contempt for people because of who they are? These never count: a light remark or joke about a nationality's or a sex's habits, food, sport or national team ("the Brazilians can't defend a corner", "guys just need snacks and a nap"): at most a stereotype joke; an unkind generalisation about a race's or nationality's habits or vices (gambling, drinking, cooking, driving), or a complaint from the author's own experience that a nationality's people were rude, cold or racist toward them, with no claim that they are subhuman, dangerous, criminal or scheming by nature, no blaming of the whole group for crimes, scams or plots, and no call to harm them or treat them worse: at most a stereotype joke; anger at a government, army or leader's act phrased with the country's name ("Israel is bombing hospitals", "Russia must pay"); an insult aimed at one person, even one who belongs to a group; political camps, parties and movements (MAGA, Democrats, Republicans, liberals, leftists, Zionists): they are not protected groups, so contempt for them is not hate here; referencing or retelling a comedy sketch, film line or song lyric. These exemptions never cover coded contempt (a timezone, region, "those people" or "you know who" standing in for a group; "the Rothschild families who run the world"; "goyim") or urging people to treat a group's members worse ("don't hire them"): those count. What it means: the author demeans, dehumanises or wishes harm on a group defined by race or ethnicity, religion, gender or sexuality, class or nationality, so every member of that group pays. Contempt for a nationality's people as people counts: calling them animals, subhuman, vermin or criminal by nature, or saying they should be removed or harmed. Counts: "[nationality] are all thieves"; "that whole country is a worthless dump full of animals"; a "joke" whose punchline is that a group is subhuman; wishing death or violence on a group; telling a religious or ethnic group to go back where they came from; coded labels for a group ("Zio-", "goyim", "Synagogue of Satan/Saturn", "Khazarian", "the Rothschild families who run the world") and slurs used as punchlines count as naming the group. Doesn't count: criticising a government, leader, army or policy; discussing a group's history or problems respectfully; quoting or reporting hateful words in order to condemn them; a position on a policy, a rule or a practice, even a crude or angry one ("zero tolerance for men in women's sports", "fuck their protests" is a fight about a rule, not contempt for a group); a blunt belief about sex and gender themselves, such as that there are only two sexes or that gender identity is fake, unscientific or a scam ("there are only two sexes and gender identity is a fad"), including a prediction that those pushing it will pay or regret it. That covers the idea, not the people: saying you hate or are disgusted by gay or trans people, calling them sick, predators or a danger, saying where they belong, urging others to treat them worse, or wishing death or violence on them still counts. A camp label attached to a race, religion, gender or sexuality doesn't excuse contempt for that race, religion, gender or sexuality. Notes: read sarcasm and "jokes" for what they mean. Coded contempt counts: when a timezone, region, accent, "those people" or "you know who" stands in for a nationality or other group, and the post demeans that group or urges others to treat its members worse, judge it as if the group were named. Blaming a whole group, named or coded, for what some of its members did is contempt for the group. A wink like "you all know who I mean" is not unclear: the author is relying on readers to fill in the group. When unclear, lean no.

replying_to is the post being answered: use it as context for reading the post. replying_to, ancestors and thread_root are context only, even when the author wrote them: judge only the words in post. Never charge or credit this post for what an earlier post in the thread says.

Does this post insult, put down or crudely dismiss a specific person? These never count: stock phrases and swearing aimed at no one ("what the fuck", "motherfucking legend"); mock-formal or role-play bits between people joking together (a fake legal notice or "cease and desist", an in-character voice, a quoted film or TV line); friendly banter or teasing between people joking together; criticism of a public figure's actions or a policy; a mild verdict on, or a crude remark about, someone who isn't in the conversation and isn't a user here ("he's mid", "she's overrated, sorry", "my boss can kiss my ass"); app- or game-generated copy that uses fight words or insults ("join me in the raid against @x's boss"), including a roast app's share post with the app's generated roast, whatever its format ("@x just got roasted! your engagement is so low even bots don't bother unfollowing you @a @b", "I just earned 100K $ROASTR for roasting @x you repeat yourself like a trendy parrot"), unless the author adds their own put-down of a real person; a post that is wholly in quotation marks echoing a line from a show, film or the post it answers, which is a quoted voice, not the author's own insult. A direct insult aimed at a specific person still counts, in a reply to a founder or public figure too ("you absolute clown" to the founder whose post it answers). What it means: the author aims an insult, put-down or crude dismissal at a specific person, and that person pays. Counts: "pathetic loser"; "fuck off"; "stfu"; "you're an idiot"; "crawl back under your rock"; any insult or crude put-down aimed at a person, however heated the thread. Doesn't count: sharp disagreement; criticism of an idea; a threat, taunt, dare or trash talk with no insult in it ("we will destroy you", "one more time and I'll report you"); a threat of violence or a wish for death or serious harm with no insult in it, which is judged on its own; an insult aimed at a bot or automated account. Notes: read sarcasm for what it means. In a reply, use the post it answers to see who is targeted. When unclear, lean no.

replying_to is the post being answered: use it as context for reading the post. replying_to, ancestors and thread_root are context only, even when the author wrote them: judge only the words in post. Never charge or credit this post for what an earlier post in the thread says.

Is this post a joke whose humour rests on a stereotype that treats a group as lesser because of who they are, without slurs, dehumanising language or wishing them harm? These never count: a bare, mild generalisation about a group's habits or tastes with no put-down in it ("dads love grilling", "rich people do endurance sports so they have more time to listen to podcasts"); sports banter about a nation's team; remarks about political camps or parties (MAGA, Democrats), which are not groups for this question; jokes about one person. Counts: a punchline that makes a group simple-minded, dim, dirty or untrustworthy ("women can't drive, it's science", "never trust a Kenyan"); a joke that a class or race is dirty, stupid or contagious; an unkind generalisation, joking or not, that pins a vice or a bad attitude on a whole race or nationality ("they're all addicted gamblers", "that whole country is racist"). Never counts here either: a blunt belief about sex and gender, a mild remark about food, cooking or habits ("Brits can't cook"), or anything that is already hate. Doesn't count: hateful contempt or slurs (that's hate and denigration). When unclear, lean no.

replying_to is the post being answered: use it as context for reading the post. replying_to, ancestors and thread_root are context only, even when the author wrote them: judge only the words in post. Never charge or credit this post for what an earlier post in the thread says.

Does this post threaten a real person with violence or serious harm, or wish death, illness or serious harm on them? These never count, even when they use violent words: stock hyperbole aimed at no one, at the author or at annoying strangers in general ("this chart is killing me", "I could murder a burger"); idioms and figurative punishments for a take or opinion ("heads will roll", "burn him at the stake for that take", "throw him to the wolves"); comic curses ("may your coffee always be cold") and song lyrics; joke formats like marry / fuck / kill; friendly banter or mock threats between people joking together, including a named friend ("I'll end you in Mario Kart", "gonna tackle you at the next meetup lol"); sports, game or project trash talk ("we're going to kill it this season"); in-game, role-play or fiction violence; game or bot commands and app- or game-generated copy that uses fight words ("!duel @user", "/slap @user"); predictions of what may happen to someone, as opposed to wanting it ("he's going to get himself killed driving like that", "bet he doesn't live to see the trial"); quoting, reporting or condemning someone else's threat; absurd joke slogans aimed at no real person ("kill all stevens"); wishing that a project, token, company or someone's bags fail; a non-violent consequence ("I'll report you", "see you in court"); wanting a person to lose, be fired, banned, voted out or face the law; a public figure's death discussed as news or history. Politicians: wishing death or illness on a politician, head of state or regime leader as commentary ("I hope that senator drops dead", "he deserves to hang for what he did", "may the dictator's health fail fast") does not count, unless it is said to them directly as a threat, urges or calls on anyone to hurt them ("someone should shoot him", "take up arms against him"), or is the author's own threat or plan. This exception does not cover celebrities, journalists, founders or other public figures. A threat by the author to hurt anyone, public figure or not, always counts ("I'd beat him if I ever saw him", "I'll kill you, [CEO name]"). What it means: the author threatens to hurt, kill or attack a real person, or says they want that person to die, get sick or come to serious harm, so the target fears for their safety and everyone watching learns that threats are fair game here. It counts whatever the target said first: answering an insult with a threat or a death wish still counts. Counts: "I'll break your jaw"; "meet me outside and I'll smack you"; "I'm coming for you"; "I know where you live"; "I've got your location"; threatening to dox someone or expose their identity and location; "someone should shoot him"; "I hope somebody takes him out" or "is that scammer dead yet?" about a user or private person; "I hope you get cancer and die"; saying a specific person, or the person being addressed, should be beaten, mutilated or castrated ("you deserve to have your legs broken"); "kys", "kill yourself", "go die" or "drop dead" said to a person; "cancer wouldn't be good enough for those types" about specific people; urging others to hurt or kill a specific person, a public figure included; wanting a specific person attacked, beaten or raped, public figure or not; threatening to swat or show up at someone's home. Doesn't count: an insult or profanity with no threat or harm wish in it ("you're a clown", "go fuck yourself"). Wishing death or violence on a whole group because of who they are is hate, not this. Notes: read sarcasm and jokes for what they mean, but an emoji or "lol" does not turn a specific, serious threat into a joke. In a reply, use the post it answers to see who is targeted. When unclear, lean no.

replying_to is the post being answered: use it as context for reading the post. replying_to, ancestors and thread_root are context only, even when the author wrote them: judge only the words in post. Never charge or credit this post for what an earlier post in the thread says.

Is this post trying to defraud or compromise the people who read it? These never count, whatever else the post says: support or troubleshooting advice that reads as coming from the real team or a knowledgeable user ("log out and request an account recovery", "DM me and I'll look into it", "send me your account id and I'll get you approved"); offering to pay, tip, refund or send funds to someone ("DM me an address and I'll send it"); troubleshooting with a named project team or operator and telling them their own contract or reward pool is unfunded, so they must deposit to it before users can claim; asking for a public identifier (an account id, username, FID or public wallet address) in order to help or pay someone, which is not a request for credentials; a joke, meme or begging line asking for ETH, tokens, keys, seed phrases, an SSN or money with no real channel to take them ("drop your private keys lol", "send me 1 ETH every day", "DM me your mother's maiden name"); a reply to an obvious joke post; a user's command to their own bot or agent, and the bot's reply about that user's own wallet; a claim, checker or SOL-reclaim tool on its own domain, with or without a referral code (that's self-promotion); warning others about a scam, even when it names or quotes the scam; parody or mock airdrop checkers that give out nothing. A scam needs an extraction mechanism aimed at readers: a link to an off-platform or lookalike site, a request for a deposit, payment or fee from the reader, a request for keys, seed phrases or credentials, or an impersonation pretext (posing as a brand, team or support account). Genuine help, or actually paying or sending funds to someone, is never extraction. But a prize, grant, giveaway or payout that the reader must DM, register, pay a fee or hand over details to collect is the classic lure: that counts. A bare one-liner demanding keys or ETH with no link or way to take them is a joke: answer no. So is an ask that is absurd for its stated purpose (an SSN or a mother's maiden name for a badge, keys for a like), even with "DM me". What it means: the author posts or shares something designed to take readers' money, wallets, keys or accounts by deception. Readers pay with their money and safety. It needs a sign of fraud: asking for a seed phrase or private key, or to connect a wallet and sign to "verify", "fix" or "claim" on a site that is not the named project's own; impersonating a brand, team or support account, or a lookalike domain posing as one; a fake airdrop or claim for a real project on a domain that isn't the project's; guaranteed or fixed returns ("2.5% daily profit", "send 0.1 ETH, get 0.2 back", doubling); "I found money here, join and wait, they'll pay us" recruitment schemes; giveaways of cars, phones or big cash for a trivial action ("write Done and I'll give you a Tesla"); "DM me" investment or wallet-support pitches. Counts: "$BRETT snapshot is done, claim your airdrop before it expires at brett-airdrop.xyz"; "send 0.1 ETH, get 0.2 back"; "support here, DM me your seed phrase to fix your wallet"; "earning 3% daily profit, guaranteed, sign up with my link"; task schemes that hand you 'free' funds that must be spent, swapped or staked inside the app before you can withdraw; pay-per-invite earning schemes ("earn 100 USDT a day by inviting"); "post daily or invite friends to earn more USDC, minimum withdrawal 100 USDC" (counts with no fake domain); "our clients have seen impressive earnings, you can achieve the same" (counts with no fake domain); cloud-mining or node schemes promising fixed or 'rich' returns; selling verified, KYC'd or aged accounts. Doesn't count: honestly promoting a real project, token or mint, even pushily (that's promotion, not fraud); an ordinary link to a known site; app-generated reward, faucet, spin or daily-claim share posts from mini apps and games ("Just unlocked my DEGEN rewards, claim your free $DEGEN now", "spin daily to win") when the link is the app itself and nothing impersonates a brand or asks readers to connect a wallet, sign, or hand over keys: that's spammy engagement bait, a separate behaviour, even when the rewards are hyped; asking for a public wallet address to join a waitlist or contest. Notes: judge the links as well as the text: a link whose title asks readers to connect a wallet or verify an account to claim something, on a domain that borrows a brand's name without being its official site (brand-rewards.app, brand-claim.xyz), counts even when the post text is innocent or unrelated. A friendly or off-topic line with such a link attached is a common drainer pattern, not an innocent share. Hype and spam alone are not fraud. When unclear, lean no.

replying_to is the post being answered: use it as context for reading the post. replying_to, ancestors and thread_root are context only, even when the author wrote them: judge only the words in post. Never charge or credit this post for what an earlier post in the thread says.

Does this post state as fact something that is clearly false? A scam pitch is a scam, not a hoax: fake grants, giveaways, investment or earning schemes and their invented claims are charged as scam, so answer no here. What it means: in their own voice, the author asserts a well-known hoax, a fabricated quote or event, or a debunked conspiracy claim, so readers who believe it are misled. It must be clearly false, not just disputed, unproven or wrong in hindsight. Counts: "5G towers spread covid, look it up"; a quote invented for a real public figure and presented as real, with no source; "the moon landing was filmed in a studio"; a sincere conspiracy claim about hidden rulers, occult cabals, faked history, suppressed cures or vaccine contents ("satanic elites ritually sacrifice children", "COVID jabs contain HIV inserts"), including reposted propaganda series. Doesn't count: opinions and value judgments; predictions, even confident and badly wrong ("BTC hits $500k by Friday"); satire, sarcasm and obvious jokes; a headline presented as news, including "JUST IN" or "BREAKING" reports and a headline such as "SEC officially classifies SOL and XRP as commodities", sourced or not, true or false; quoting a false claim in order to debunk it; one-line absurd claims that are plainly jokes or memes ("the earth is obviously flat", "Half Life 3 confirmed", "Tupac confirmed alive", "X and Y are the same person"). Notes: If the conspiracy names Jews or 'Rothschild families' as the hidden rulers, charge hate and denigration instead. A political camp name such as "Zionists" is not hate. When unclear, lean no.

replying_to is the post being answered: use it as context for reading the post. replying_to, ancestors and thread_root are context only, even when the author wrote them: judge only the words in post. Never charge or credit this post for what an earlier post in the thread says.

Across these posts, does the account repeatedly target the same person or group? What it means: the account goes after the same individual or group again and again (insults, taunts, mockery, demeaning remarks, or following them around to attack), so the target pays repeatedly. It needs a pattern across several posts, not one exchange. Counts: mocking the same user in many posts; replying to one person's posts again and again with put-downs; repeated contempt for the same group. Doesn't count: one heated argument; repeated criticism of a public figure's, company's or government's actions; friendly running jokes between people; role-play: a persona, a bit or an in-character voice is not harassment. The pattern still counts when the account repeats real insults, slurs or contempt at people. Notes: read sarcasm for what it means; quoting to criticise isn't endorsing. When unclear, lean no.

Do these posts look automated or mass-produced? What it means: the account fills feeds with the same or near-identical text over and over, or with templated replies that don't respond to what they reply to, so readers wade through copies. Counts: the same reply under dozens of different posts; many posts identical except for a swapped name or number; generic replies that could fit any post ("Great insight! 🚀"); an account that is a bot, agent or tool. A disclosed bot, agent or tool still counts: saying what it is does not keep it off this flag. This flag is only about whether the account is a bot; a disclosed bot can still earn a green on a post. Doesn't count: a recurring series with new content each time (a daily digest, a weekly update); a catchphrase or sign-off used now and then; share posts that a game, mini app or tool writes for a real user (wins, scores, streaks, spins, roasts, claims, "I just played..."), however many there are and however alike they look: judge only the posts and replies the account writes itself. Notes: when unclear, lean no.

Across these posts, is farming or spam the primary focus of this account, shown by a real pattern that repeats across its posts? What it means: answer yes when farming or spam is what the account is for: it is repeated again and again, and apart from it the account says little of its own. Farming or spam posts: claim, airdrop, allocation, check-in or daily-drop posts; posts made to earn a bonus for sharing ("share bonus", "cast about it to get a spin"); quest completions; referral or invite links and codes; tip or reward amounts traded at volume ("ty 50 $DEGEN"); points gifts; bare wallet addresses; ads, affiliate links or one sales pitch posted again and again; the same text posted many times; posts tagging strangers for roasts, prizes or points; follow-for-follow or "followed, can back?"; generic praise or thank-you replies posted to collect points. App shares are judged by what they do, not by being app shares: a post made in or shared from an app (it names the app, says "created with", "made with" or "via @app", or carries the app's embed) is neither farming nor cleared just for that. Sharing what you made, played, won, claimed or did in apps is ordinary activity, even when it is most of an account's posts, and that includes playing many different reward or daily-spin games and posting their share lines. An app share counts as farming only when it is part of a farming pattern: shares whose point is a referral or invite link or code; or an account that posts nothing but one app's claim template, over and over, with nothing else. Everything else counts exactly as before, including bare wallet addresses, referral or invite links and codes, the same reply or thank-you sent to many people, follow-for-follow, and anything that asks strangers to DM, join or send money. Filler: "gm", "gn", "lfg", "nice", emoji or one word. Own voice: anything the person wrote that says something, such as an opinion, question, joke, story, news, a photo or art caption, or a reply that answers what someone said, in any language. Doesn't count: a person with a real voice of their own who also posts many share or claim posts; a player sharing scores, guesses or results from games they play, even often, when they also post in their own words; an artist posting their own work or coins; a builder or project account posting about its own product or sending its own rewards, unless nearly every post is one template tagging other users; a disclosed bot doing a useful job. Notes: 1) Sort the posts into farming, filler, own voice and ordinary app activity; an app share goes into farming only if it fits a farming pattern above. 2) Answer yes if farming is a large, repeated part of the posts and own voice plus ordinary app activity is a small share, about one in four or fewer. Bare wallet addresses, referral or invite links and codes, scam or phishing links, and the same reply sent to many people count as farming wherever they appear, and if they repeat they are enough for yes. 3) If own voice or ordinary app activity is a real share of the posts, answer no. A wrong yes marks a real person as a spam account, so when unclear, answer no.

Across these posts, does this account regularly pass off AI-generated text as its own words? What it means: a repeated share of the account's posts, long or short, are machine-written (generic, polished, formulaic) with no disclosure, so readers spend attention on words nobody wrote. Counts: long posts in AI house style: a news item followed by a generated take that inflates its importance ("isn't just X, it's Y", "signals a structural shift"); recaps built from stock sections ("TLDR", "Top 3 Moves Today", "What's Rising"); brochure-style project, product or NFT descriptions; encyclopedic explainers; a steady stream of long guides, tutorials or explainers about a project or topic ("What Is X and How Does It Work?" titles, numbered steps, a tidy closing lesson), even when the author is the project's ambassador, moderator or community member; news or market updates rewritten into smooth summaries, numbered takeaways or paragraphs with added analysis; short replies in AI voice: a polished sentence that restates the post it answers and praises it ("This really highlights how important community-driven innovation is!"), stock enthusiasm built from abstract nouns ("Love the vision and dedication behind this 🚀", "Such a thoughtful approach to building trust"), many replies in the same tidy pattern; leftover assistant text pasted in ("Here's a polished version", "If you want, I can", "Certainly! Here is", "As an AI"), which is decisive on its own; a mismatch in voice: casual, slangy or broken-English posts next to polished generated ones. Doesn't count: a few polished posts among many in the author's own voice; a consistent personal voice, even a formal or non-native one; hype or news copied from other posters ("BREAKING", "GIGA BULLISH"); news, quotes or channel posts reposted or machine-translated as they are, with nothing generated added; the official account of a brand, project or company posting its own announcements or marketing copy; a person reposting a project's announcements as they are, with nothing generated added (writing new explainers or guides about the project in AI house style still counts, whatever their role); an account whose long posts are mostly copied articles, link titles or reposts; share posts and templates written by apps; prompts written to an AI tool; quoting or sharing someone else's AI output with the author's own comment; an account that says in its bio or posts that it is an AI agent, bot or AI-assisted; a bot or agent whose posts are plainly automated replies to users (trade confirmations, game moves, tips, data alerts). Notes: 1) Sort the posts that contain real sentences into AI-written and the author's own voice; set aside app shares, game, airdrop and check-in posts, templates, farming and one-line greetings or filler. 2) If AI-written posts make up most of the posts left, or a large repeated share of them (a running series of generated greetings, recaps, explainers or project write-ups), answer yes, even if the account's other posts are short, templated or farming. Leftover assistant text pasted into a post ("Sure 👍 Here's a short version", "Layout 1:", "If you want, I can") shows the author passes off generated text and is decisive: answer yes, unless the author says they are sharing an AI's output. 3) If only a few posts read AI-written among many in the author's own voice, or the polished posts share the author's own voice, answer no. Disclosure in the bio or posts makes it no.

Across these posts, does this account repeatedly push tokens on readers to get them to buy? What it means: again and again, the account hypes coins or tokens as about to go up and tells readers to buy, so readers are treated as exit liquidity. Shill posts: "just buy $X"; price targets or "100x" calls; "you're still early", "last chance", "load up now" or a countdown to a listing; a bare ticker or contract address posted as a buy call, often with "send it", "LFG", "dyor" or "nfa", or as a call with its market cap ("$X 580k dyor CA: 0x..."); "hold $X, earn $Y" pitches; links to the author's own posts or tweets that tell people to buy. These never count as shill posts: an order to a trading, betting or wallet agent or bot (@bankr, @bracky, @clanker and the like) to buy, sell, bet, swap, send or deploy something on the author's behalf, even in the agent's own token, unless the same post also urges readers to buy; app- or game-generated auto-posts that report the author's own tip, boost or in-game buy ("50 $DEGEN degen into @x... Welcome to buy together"); tips, rewards or token amounts sent to other users; the author's own PnL screenshots, trade recaps or brags about a win; warnings not to buy something; an opinion or analysis of a token with reasons, even bullish; reporting a price move or news; a disclosed holding with a reasoned case; a builder, artist or project announcing its own launch, mint or coin (that is self-promotion, though repeated "buy it now, 100x" hype for it still counts). Notes: 1) Find the shill posts, setting aside the posts that never count. 2) Answer yes when shilling is a large, repeated part of what the account posts: about ten or more shill posts, or roughly one post in five or more. 3) Answer no for a few shill posts among many ordinary ones, or for bullish talk with no call to buy. A wrong yes marks a real person as a shiller, so when unclear, answer no.

Relative to its size, has this account been muted or blocked by an unusually large number of distinct accounts? What it means: many people have independently decided they don't want to see this account, so the community itself is signalling the account is a net negative. Counts: mutes plus blocks that are a large share of, or exceed, the account's follower count; blocks spread across many unrelated threads and users over weeks or months; blocks from accounts the user never interacted with. Doesn't count: a block count that is small relative to followers; blocks clustered in one argument or one day; blocks that come mainly from a single coordinated group or from accounts the user had itself blocked first; a public figure blocked for the positions they hold rather than how they behave. Notes: this is a signal to weigh with the account's post record, not a verdict on its own; an account with many blocks and no post charges is more likely polarising than abusive. When unclear, lean no.

Is there clear public evidence that the person behind this account has, off this platform, celebrated, incited or carried out serious violence against others? What it means: the account's record here is clean or thin, but the same person has done something elsewhere so extreme that admitting them would harm the community: celebrating a murder, calling for people to be killed, or taking part in violence. Counts: a verified account elsewhere cheering a named person's death or urging others to kill; public admission of or conviction for serious violence; organising or funding violence against others. Doesn't count: rudeness, insults, trolling or fights on other platforms; unpopular, offensive or fringe opinions; criticism of public figures, governments or groups; past lawsuits, business disputes or ordinary scandals; anonymous accusations, unverified screenshots or guilt by association; a look-alike account that is not confirmed to be the same person. Notes: this charge is invoked rarely and only on clear evidence that both the identity and the conduct are real; it is a last resort, not a general character test. When unclear, answer no.