Henry Zhang

this is the real problem — most tools optimize for the writing, not the judgment of what's worth writing in the first place. the second-order thinking is deciding what tone fits the moment, not just picking one from a dropdown.the real problem is most tools optimize for the writing, not the judgment of what is worth writing. the second-order thinking is deciding what tone fits the moment, not just picking one from a preset.

most AI tools give you one dial. formal. casual. professional. friendly. that's not how writing actually works. a contrarian take needs different cadence than a founder reflection. a market breakdown needs different hooks than a personal story. a reply to a competitor needs different energy than a thread about lessons learned. Auden inside @voicemoat runs on 13 tone presets. all trained on how you actually write. witty. contrarian. punchy. reflective. educational. provocative. narrative. hype. chill. conversational. thoughtful. question. story. same raw thought. completely different output depending on the moment. #BuildInPublic #IndieHackers #BuildWithAI

@zoiroff77 @OlegBevz the founder bottleneck hits hard when every reply needs to sound like you - not like the tool you used to write it. the hard part is deciding what context matters, not having the words.

Same logic applies to sales.The most expensive mistake is the founder writing every proposal themselves. The fix mirrors yours: 30 minutes a week training the team + sharing context from won deals, everything else happens without the founder. Founder bottleneck in content kills reach. In sales, it kills the business multiple

1/5 Founders who write their own content get around 4x more engagement than founders who fully delegate it. That gap doesn't close with better writers or bigger budgets

this is exactly right - and the real issue is the dial assumes you know what tone you need before you know what you want to say. the harder question is always: what kind of reply does this moment actually need?the real problem with one dial is it assumes you know what tone you need before you know what you want to say. the harder question is always: what does this specific moment actually need?

@LilithLovett lol i feel like we're going to see a lot of these 'mechanical turk' moments as companies rush to ship humanoid tech. the internet filters this stuff out pretty quick though

Why are they only doing 8-hr shifts if they're robots? Why did that robot touch his head? -...It's Indians pretending to be robots again isn't it?

@DeepDishEnjoyer honestly i feel like the internet just loves to hate on anyone putting themselves out there lol, it's not even that long

one succinct observation is that nicholas decker thinks his experience is worthy of a two-page resume

@Andercot it's basically just the glasshole effect all over again. people really underestimated how much 'looking normal' matters for consumer tech vs just having the best specs

It's probably as simple as "you don't look cool wearing it" Zucks RayBans are possibly the only way forward

@haridigresses it's basically just adsense for the series a crowd lol. everyone is just trading the same vc dollars in a circle

In case you're wondering, this is the stage of the market we're at.

@bhalligan the hard part of board meetings is not writing the update — it's deciding which context actually matters to the board. that filtering work is invisible but it's the real leverage.

Every board meeting 1. We are growing v fast 2. We need to fix gross margins 3. We’re behind on enterprise reps 4. We’re moving to credits pricing 5. We have one pesky competitor 6. We’re very worried re Anthropic If this is you, you have a lot of company.

@daltonbrewer The hard part is not writing — it is having something worth saying. AI can mirror the surface of your words, but it has no idea why you actually said it. That's why it always sounds off. The gap isn't polish, it's situation.

AI replies on X are super annoying It’s all the same generic crap, essentially mimicking the post Please don’t be that person…

@ferobujnovsky the hard part is not writing it is deciding what context matters

Wrote to ChatGPT for a quick grammar check - my usual routine when I’m not sure about my English. What does it reply? “Yeah, it works, but it sounds a bit generic and slightly AI-ish. The idea is good, I’d just make it shorter and sharper.” AI-ish???? What do you mean?? Cancelling the subscription 😂

@MatanHazanov the no-context pitch is a founder hygiene problem. investors need to care about the problem before they can care about your ask.

This was the first message I received from a founder. I genuinely didn’t know what to do with it. Needs $1M immediately. Ideally yesterday. Another $11M right after. $10B valuation within a year. No product. No details. No context. Just... confidence. I’m still not sure who this is supposed to work on.

@trikcode hello world test

i asked claude opus 4.7 to refactor a large codebase. 68 minutes, millions of tokens burned - it finished nothing worked. app completely broken

@JustSpyda_ this is the part nobody measures: the energy cost per real conversation. 200 low-effort replies feel productive because they create activity, but they do not create relationships. the real metric is how many replies you would actually stand behind if the person replied back.

200 replies. zero conversations. you're calling that engagement. Nearly every founder and community manager thinks a dying community goes quiet first, but it actually gets loud before it ever goes silent. You start seeing more posts, more announcements, more contests, more "gm" threads with 200 replies and zero substance. Your dashboard says engagement is up 40% month over month and you screenshot it for your investor update. But what you're actually looking at is a community trying to resuscitate itself by flooding the surface with activity because the foundation underneath already cracked. This is the part people choose to ignore; the volume is not momentum, it's the immune response of a system that lost its reason to exist, and it's the deadliest phase because it looks like progress on every metric you're tracking. Here's what's actually happening while your numbers climb: the quality of participation is collapsing in real time. Scroll through your chat and count how many messages are actual conversations versus how many are one-word reactions, emoji chains, and replies that don't reference the message above them. You'll find that 80% of your "engagement" is performative. People aren't speaking because they have something to say, they're speaking because the system rewarded them for showing up; a quest here, a role upgrade there, maybe some points that convert to nothing. That's decay, and it compounds fast. The instinct when engagement drops is always the same: produce more content, ship another campaign, launch a twitter space, run a giveaway. I've watched community managers triple their output in a single month and wonder why retention kept falling. They were treating engagement like a math problem where more input equals more output, when engagement is actually a side effect of meaning, and you can't manufacture meaning with a content calendar. Think about the last community you actually wanted to open every day. It probably wasn't the busiest one. It was the one where you recognized names, where someone remembered your last comment and built on it, where decisions felt connected to what members actually said. That's gravity. People return to communities that have clear identity, shared language, and visible proof that participation changes outcomes. When those three things erode, the only thing left to fill the gap is empty volume. And empty volume doesn't just mask the problem, it accelerates it by exhausting the exact people holding your community together. Your strongest contributors are the first to leave. They're the ones who wrote the thoughtful posts, answered questions before mods got to them, and enforced norms just by showing up. They notice before anyone else that conversations stopped compounding, that feedback gets acknowledged but never acted on, that the space feels more like a broadcast channel than a room they belong in. They don't announce they're leaving, they just stop opening the app; and when they disappear, the volume gets louder because the social pressure they created, the unwritten rules about what a good post looks like, how disagreements get handled, what's worth discussing, vanishes with them. If you want to stop this, you have to stop treating chat volume as a health metric and start reading it as a symptom sheet. A community that's actually working doesn't feel busy, it feels intentional. Conversations exist because someone needed an answer or wanted to build on an idea, not because a prompt told them to post. Members reference each other by name. Norms get enforced by the room, not just the moderator badge, and new members learn how to behave by watching what gets respected, not by reading a welcome doc they'll forget in ten minutes.

@Motion_Viz the re-explain part is the real cost. not just the minutes, but the mental reload every time you context-switch. you basically pay a cognitive tax 15x a day.

the biggest time killer in seo isn't keyword research it's context-switching chatgpt for briefs → ahrefs for data → back to chatgpt to re-explain everything 15 times a day. on desktop it's annoying. on mobile it's unusable client calls about a rankings drop and by the time you've copy-pasted enough data into an ai to make sense of it, the call is over the tools aren't the problem. the workflow is been thinking about what an ai-native seo workflow actually looks like — one where the context stays persistent and the data pulls happen inside the conversation might build something around this. curious if anyone else is drowning in tab-switching hell or if it's just me

@harshilmathur thinking about this a lot with blink. we're trying to use ai to give people superpowers rather than just automating away the human element. if you're curious about the builder side of this, check out github.com/getblink/blink

Companies that are just using AI to replace people are underestimating the power of both.

@ZackKorman honestly it feels like models are just trained to be polite critics by default. the "meaningless critique" thing is such a huge part of the ai sycophancy problem tho

People talk about the AI sycophancy problem, but the inverse problem is meaningless critique. Upload a real Rothko, say you made a reproduction, and ask how you did. Gemini 3.1 Pro is the only model that consistently says "yea that's good"

Deadlines create the constraint that forces decision-making. Without them you optimize infinitely. The trick is choosing which deadlines to accept and which to ignore—saying no to fake urgency is as important as respecting real ones.

@alexwtlf volume can compensate for conversion rates for a bit, but it feels like a race to the bottom. eventually the 'human' signal is the only thing that will actually convert.

what happens when every startup has AI influencers, AI marketers, and infinite content generation? What becomes the bottleneck then?

@NayanUnfiltered this is exactly what we're thinking about with blink. avoiding those ai tells like em dashes and hedging is basically the whole game now if you want people to actually read the content.

1/8 Every reply: - Under 280 characters (hard limit) - Sounds human, not like ChatGPT - Engages with the actual post content - No generic "Great take!" openers - No AI tells (em-dashes, hedging, etc) The model is told: "Write like someone with a real point of view. Confident. Direct. Spicy when warranted."

I Built ReplyForge : a pure local AI agent that generates sharp banger replies in seconds. github.com/ZeusSpider007/… If you've ever been frustrated by generic AI reply suggestions or wanted to own your tooling, this might scratch that itch. No cloud. No API keys. No telemetry. Just your Mac + Ollama + a language model running entirely offline. Paste any tweet → get 3 ready-to-post replies (professional, bold, witty). Follow the Thread Below 🧵 #LocalAI #Ollama #OfflineAI #OpenSourceAI #AI

@Awwaluuuu lfg! it's in closed beta testing right now. i'm super curious about your use-case, dmed you. in the meantime, share blink w ur friends!

@thehenryinsf I already joined the beta when will it launch ? Is there access to testing?

I aim for 100 intentional replies an hour. Not random noise, replies that pull people back to the account. People call it being a 'reply guy.' I call it driving 33 Million Impressions (+4,381%) and 10.5K conversations in a single month. Authenticity isn't a strategy; it's a requirement. If you don't love talking to people, the audience feels it. SMM isn't scheduling posts; it's reading the room.