REJ

Twitter thread or long tweet

cmpcqnfbs00724o3llizv60tk · 𝕊𝔸𝔸𝔻 @saadamrun

User

𝕊𝔸𝔸𝔻

@saadamrun

Submitted

2026-05-19 14:38:45 UTC

Cohort

Cohort 1

closed

Reward base

7 points

Attempts

1 / 3

base 3

Attempt 1 of 3

Submission URL

https://x.com/SaadAmmrun/status/2056694613470056625 ↗

Validator · manual

manual: needs-review

Review result

Reviewed at

2026-05-20 07:26:51 UTC

Reviewer

system

Comment / reason

this is the most gpt-shaped thread in the batch. the angle is fine. "verified record beats disappearing discord roles" lines up with what internpass does. but the prose has every AI tell in the playbook stacked on top of itself. running through the patterns so you can clock them yourself next time. zero contractions across the whole thread. "i did not expect", "it is not about", "it is about", "there is no", "you are active", "i have put my work on record", "with appreciation", "and to internpass". humans contract. when a piece never does, the model fingerprint is on it. start saying "didn't / isn't / there's / you're / i've" and 30% of the AI smell evaporates immediately. essay titles on tweets. "acknowledgment of verification", "the structural problem in web3 syndicates", "formal invitation for web3 professionals*". op-ed framing, not tweet framing. tweets work on first-line hooks, not headers. drop all three titles. lead with the strongest sentence of each tweet instead. antithesis structures stacked. "it is not about ego. it is about having my contributions formally acknowledged." / "not temporary discord roles. not chat logs that disappear." / "not a vanity badge, a professional record." three "not X, Y" moves in one thread. it's gpt's favourite rhetorical device because it sounds decisive without committing. pick one antithesis, cut the other two. triadic lists everywhere. "you contribute consistently. the project slows down or dissolves. your work history vanishes with it." / "i was there. i contributed. this is what i did." / "manually reviewed proofs. on-record roles. verified by @sndcapp." the model groups things in threes because three feels balanced. when every section does it, the cadence becomes the signature. break some into pairs, fold some into prose. corporate vocabulary. "formally acknowledged", "verification standard", "professional record", "the reality is different", "quietly but effectively", "changed my perspective". these are words from quarterly reports, not from someone tweeting. test each phrase. if you wouldn't say it aloud to a friend, swap it for what you'd say. "we often speak of web3 as the future of work". distancing essayist opener. you're not the world economic forum. nobody on crypto twitter "speaks of". try "everyone calls web3 the future of work. for most contributors it isn't." same idea, your voice. "actually need". drop "actually" — pure hedge. what to fix on next pass: 1. add contractions everywhere they fit — biggest single change 2. cut all three tweet titles, lead with the strongest line 3. keep one antithesis structure, cut the other two 4. break the period-triplet rhythm, let pairs and longer sentences appear 5. swap corporate phrasings (formally acknowledged / verification standard / professional record) for plain english 6. rewrite the "we often speak of" opener as something you'd actually say the underlying observation is strong and the thread structure works. once the model accent comes out of the prose, you've got a real person making a point worth listening to, since this task is biweekly, even though is good in visuals, i have to reject it, i believe you can do much better! looking forward to see another take

Submission rejected · no further actions

"this is the most gpt-shaped thread in the batch. the angle is fine. "verified record beats disappearing discord roles" lines up with what internpass does. but the prose has every AI tell in the playbook stacked on top of itself. running through the patterns so you can clock them yourself next time. zero contractions across the whole thread. "i did not expect", "it is not about", "it is about", "there is no", "you are active", "i have put my work on record", "with appreciation", "and to internpass". humans contract. when a piece never does, the model fingerprint is on it. start saying "didn't / isn't / there's / you're / i've" and 30% of the AI smell evaporates immediately. essay titles on tweets. "acknowledgment of verification", "the structural problem in web3 syndicates", "formal invitation for web3 professionals*". op-ed framing, not tweet framing. tweets work on first-line hooks, not headers. drop all three titles. lead with the strongest sentence of each tweet instead. antithesis structures stacked. "it is not about ego. it is about having my contributions formally acknowledged." / "not temporary discord roles. not chat logs that disappear." / "not a vanity badge, a professional record." three "not X, Y" moves in one thread. it's gpt's favourite rhetorical device because it sounds decisive without committing. pick one antithesis, cut the other two. triadic lists everywhere. "you contribute consistently. the project slows down or dissolves. your work history vanishes with it." / "i was there. i contributed. this is what i did." / "manually reviewed proofs. on-record roles. verified by @sndcapp." the model groups things in threes because three feels balanced. when every section does it, the cadence becomes the signature. break some into pairs, fold some into prose. corporate vocabulary. "formally acknowledged", "verification standard", "professional record", "the reality is different", "quietly but effectively", "changed my perspective". these are words from quarterly reports, not from someone tweeting. test each phrase. if you wouldn't say it aloud to a friend, swap it for what you'd say. "we often speak of web3 as the future of work". distancing essayist opener. you're not the world economic forum. nobody on crypto twitter "speaks of". try "everyone calls web3 the future of work. for most contributors it isn't." same idea, your voice. "actually need". drop "actually" — pure hedge. what to fix on next pass: 1. add contractions everywhere they fit — biggest single change 2. cut all three tweet titles, lead with the strongest line 3. keep one antithesis structure, cut the other two 4. break the period-triplet rhythm, let pairs and longer sentences appear 5. swap corporate phrasings (formally acknowledged / verification standard / professional record) for plain english 6. rewrite the "we often speak of" opener as something you'd actually say the underlying observation is strong and the thread structure works. once the model accent comes out of the prose, you've got a real person making a point worth listening to, since this task is biweekly, even though is good in visuals, i have to reject it, i believe you can do much better! looking forward to see another take"