Founder of @Tonebird_ai, @tryclico(Product Hunt #1) Also a standup comedian haha, ex Sequoia China,TikTok
San Jose, CA
Joined October 2025
- Tweets3.9K
- Following888
- Followers1.5K
- Likes1.1K
Pinned Tweet
I raised $13M pitching a proactive agent.
The demos impressed rooms. But the product wasn’t becoming part of anyone’s day. that was painful to admit.
I had to let go of (my ego) and needing the idea to sound fancy. i stopped asking, “What can’t Codex or @bot do?” and started paying attention to what people were struggling with in their everyday work.
Messages kept piling up across WhatsApp, iMessage, and email.
Keeping up meant digging through old threads, switching apps, and worrying about tones.
Something so small was taking up so much time,energy, and headspace(for me, messages and scheduling eating up 50%+ of my day).
So I built ToneBird for it. Launching today ↓
woooooo!
The scariest thing this Halloween? The message you're ghosting. 👻
Ghosting Awards 🏆 Drop the scariest message you ever got, ghosted or not. Prizes:
👻 Scariest Message → smart ring
💀 Most Cursed → giant keycap light
🕯️ Most Relatable → keycap set
🎲 Random draw → keycap set
To enter:
1. Follow @Tonebird_ai
2. Repost this
3. Reply with the message (blur the names). Your boss's Friday "quick question". Your ex's 2am "u up?"
🎁 Plus: everyone gets 1 month of ToneBird Pro, free ($40 value) ↓
Ends Oct 31. Winners Nov 1.
managing an AI-native company got brutally simple. two numbers.
how much code you shipped that the company didn't have yesterday. how much attention you brought in.
zero on both and you're not on the team, you're on the payroll.
don't tell me what you've been thinking about. if an idea can't become an experiment, it's just drama in your head. make it something I can attribute, trace and test, or keep it to yourself.
desktop agents are dead. @cluely, @LittlebirdAI and my startup all bet on them.
I did the math on Cluely's consumer revenue. it's a lot smaller than what they said.
so I wrote the autopsy. three walls, and every desktop agent I know hit all of them.
1/ the install wall. every permission it needs (screen recording, Accessibility, then the work laptop policy) takes another bite out of who can even use it.
2/ the money wall. ~494k monthly visits to Cluely, my guess at conversion and churn, and what's left once the demo meets a credit card.
3/ the job wall. name one thing only a desktop agent can do a web agent can't do it, @ChatGPT can't do it.
I kept telling myself people just weren't ready for AI. that line was hiding all three.
since someone’s gonna ask: here’s the $1.6M math.
494k monthly visits (Similarweb, Sept 2026)
× 0.2% visit → paying account = ~988 new subs/month
÷ 15% monthly churn = ~6,600 subs at steady state
× $20/mo × 12 = ~$1.6M ARR
0.2% and 15% are my assumptions. typical-ish consumer SaaS. enterprise deals not included.
argue with the inputs, not the shape. if you have better numbers I’ll update the post.
I'm deleting Instinct. Used it next to Muse for a month and that was more than enough.
When two products do pretty much the same thing I just go with the faster one. Muse replies faster than anything else I've tried, and for normal everyday stuff that isn't that hard, speed is a huge part of it. I can do five rounds with Muse and Instinct still hasn't finished its first thought.
This month all of SF was losing its mind over Instinct. Two months ago it was Town AI. Next month it'll be something else, then the month after.
Same playbook every time. Someone posts "this product is sick," the founder goes on the podcast tour, everyone here hypes it, VCs pile in, and the valuation goes up while the product is still thinking lol
I'm done being the mark. I've fallen for this exact thing way too many times.
Muse just saved my life. Okay, my LinkedIn, but still.
Muse might've made me the first person ever to get a permanently banned LinkedIn account back.
End of September my post introducing @Tonebird_ai blew up, and the next day my account was banned.
I appealed, and by mid-October the answer was final. No way back.
So I let Muse go full gremlin. It wrote 50+ emails for me, pleading my case to anyone at LinkedIn who might listen.
And then one of them actually landed in the CEO's inbox !!!😱
The CEO passed it to Trust & Safety, they hopped on a call with me, and my account is back. I still can't believe that just happened.
Muse carried, and I swear I'm using it way more from now on. Claude just kept telling me not to stress and get some sleep. Sleep my ass.
And huge thanks to LinkedIn for taking a second look and clearing my name. Really great company!!
Social media can really mess with your head. Part of why I started taking X seriously is I want my own voice in here, instead of getting dragged along by whatever's trending. Scrolling X used to give me anxiety. Now I look at everything and still make up my own mind. Most of it is noise though, and the real signal is almost always in your users and in the boring basics.
every startup is playing the lag game: pick a big company and bet you can build a business before they get their shit together.
we picked Apple. ToneBird is us trying to build a personal assistant for messages before they get it right. relatively forgiving opponent bc they move slowly.
@wisprflow picked Apple too. if Apple ships better dictation, I don't think there's any room left for Wispr Flow.
some startups picked @bot . I don't envy them.
Granola went up against Zoom's built-in AI meeting notes and still hit a $1.5bn valuation.
AI lets us ship faster. unfortunately, Apple gets to use it too.
so we're betting AI won't fix the endless coordination or the boomers who can't get the important stuff shipped.
as a user, I'd love Apple to get this right.
as a founder, I want to use the lag to find what users still have to do by hand. by the time Apple nails replies, we should already be building the relationship OS.
give them something else to catch up on.
that's the lag game. keep playing
@glean vs Microsoft: working
@ElevenLabs vs OpenAI: working
Humane AI Pin (@Humane) vs the iPhone: didn't work
@MeetGamma vs Microsoft: working
@perplexity_ai vs Google: working
@gethuxe vs NotebookLM: didn't work
@relay vs Zapier: didn't work
Instinct (@noahrshinn) vs Meta's Muse: ?
founders, did I get your opponent right? and how's the game actually going?
If you’re building a consumer app, leave San Francisco. As soon as you can.
Because in SF, nothing is worth building. Everything’s “just a prompt.”
@karpathy said it last week: most of the world has barely touched an LLM, and the rest use it as a slightly better Google.
SF doesn’t see them. It doesn’t see the real needs those people have. It’s too busy with beautiful narratives, airtight logic, and top-down strategy decks that all sound right and all point at the same three ideas.
The decks are in here. The users are out there. Pick one.
"Just a prompt/skill" is the most expensive lie in AI in SF right now.
A friend's going all in on UGC content analysis. I almost told him not to. It's just a prompt, right? Nothing to build.
Then he asked what I use for it, and I couldn't name one product. Not one I've seen. One I use. Every day. For money.
What I had used was Claude. My logic was the same as everyone's: anyone with a prompt gets 80% of the way there in an afternoon. So I'd ask it to go find the videos, break them down, and it would. Structure, hook, a score. Just like Exolyt, just like every "paste a link, get a 7/10" tool out there. And I'd read it, nod, and close the tab. Not a single piece of content we've made came from one of those breakdowns. They're fun to look at. They don't change what we ship.
Could I get there with a better prompt? Probably. But a prompt that actually works here, a skill that knows which details matter, isn't something you write in an afternoon. It's weeks of running it, reading the output, finding what it missed, fixing it, running it again. The "80% in an afternoon" was the lie. It's the 80% nobody needs.
That's when it flipped for me. Look at this category for ten seconds and it's AI's comfort zone, too shallow to build on. Look at it for twenty minutes and there's a huge, unserved need sitting in plain sight. And it's not one category. Content analysis. Viral post scraping. Competitor research. UGC insights. Every one of them has real demand, real budgets, and nothing that owns it.
So here's the strange picture.
Users have the need. I'm one of them. Founders are going after it. My friend is one of them. And somehow almost nothing in these categories gets used, gets bought, or makes money.
Meanwhile I can't find a product that solves my problem, so I do what everyone does: write my own skills on top of a general agent. Those skills are shallow. They skip the details that matter, so I'm back tweaking them every single day. Prompting is convenient. Getting the output I actually want is not. It eats more of my time than a product ever should.
Demand on one side. Builders on the other. Nothing in the middle that works.
Why?
Two reasons, and neither is the model.
First, thinking. Everyone's using AI for the same things. Same queries, same tools, wildly different results. My friend explained why. "Break down this script" sounds simple until you ask: by what? Storyline? Visual style? The stickers? The motion when the product appears? The model will do some of this on its own, but the clearer you've thought it through before you prompt, the better it executes. That's the whole gap.
Second, data, and I'm sure this one won't get fixed. TikTok cut the search function out of its own Creative Center in 2024 after researchers used it to study the platform. The one public trend tool, and they pulled it. I used to work at TikTok. I know what's on the priority list. The API platform has never been near the top, and for most Chinese companies it never will be. Every analysis product downstream of it is built on sand.
So the work is in the details. Hundreds of small decisions nobody sees. Which is what I told him: don't go big this time. Find one wedge so small it looks embarrassing, and go deeper on it than anyone has bothered to.
And if you actually do that work, here's the trap.
You spend weeks on details nobody sees.
When you finish, you have to prove you beat a user who just prompted a general agent. The same user who, like me, is quietly unhappy with their own output but will never admit it on a sales call.
Then a new model drops, and half your skills, structure and fine-tuning look pointless.
Then you launch. Into a feed where "built with AI in a day" products are everywhere. I don't even open them anymore. One glance, swipe. Nine out of ten are slop, so the reflex is right. But the reflex has a cost. The products built with real care get swiped past too. Not because they're bad. Because they're buried under more noise than the internet has ever had, and I'm one of the people scrolling.
I know founders who built genuinely good things in these categories this year. Almost none of them got found. They didn't fail at building. They failed at being seen.
That's the one thing I don't see getting fixed: distribution. For Chinese teams going global especially, it's not something we're born with. We have to earn it, and most of us haven't yet.
So this is the market right now.
Users with a real need who can't find a product, so they prompt, and aren't satisfied, and tune, and tune again.
Builders who did the hard work, invisible.
And a whole set of categories everyone wrote off as "just a prompt" that are, in fact, wide open.
This is the best era for people like us. It's also the worst.
The hard part was never the idea.
"Just a prompt/skill" is the most expensive lie in AI in SF right now.
A friend's going all in on UGC content analysis. I almost told him not to. It's just a prompt, right? Nothing to build.
Then he asked what I use for it, and I couldn't name one product. Not one I've seen. One I use. Every day. For money.
What I had used was Claude. My logic was the same as everyone's: anyone with a prompt gets 80% of the way there in an afternoon. So I'd ask it to go find the videos, break them down, and it would. Structure, hook, a score. Just like Exolyt, just like every "paste a link, get a 7/10" tool out there. And I'd read it, nod, and close the tab. Not a single piece of content we've made came from one of those breakdowns. They're fun to look at. They don't change what we ship.
Could I get there with a better prompt? Probably. But a prompt that actually works here, a skill that knows which details matter, isn't something you write in an afternoon. It's weeks of running it, reading the output, finding what it missed, fixing it, running it again. The "80% in an afternoon" was the lie. It's the 80% nobody needs.
That's when it flipped for me. Look at this category for ten seconds and it's AI's comfort zone, too shallow to build on. Look at it for twenty minutes and there's a huge, unserved need sitting in plain sight. And it's not one category. Content analysis. Viral post scraping. Competitor research. UGC insights. Every one of them has real demand, real budgets, and nothing that owns it.
So here's the strange picture.
Users have the need. I'm one of them. Founders are going after it. My friend is one of them. And somehow almost nothing in these categories gets used, gets bought, or makes money.
Meanwhile I can't find a product that solves my problem, so I do what everyone does: write my own skills on top of a general agent. Those skills are shallow. They skip the details that matter, so I'm back tweaking them every single day. Prompting is convenient. Getting the output I actually want is not. It eats more of my time than a product ever should.
Demand on one side. Builders on the other. Nothing in the middle that works.
Why?
Two reasons, and neither is the model.
First, thinking. Everyone's using AI for the same things. Same queries, same tools, wildly different results. My friend explained why. "Break down this script" sounds simple until you ask: by what? Storyline? Visual style? The stickers? The motion when the product appears? The model will do some of this on its own, but the clearer you've thought it through before you prompt, the better it executes. That's the whole gap.
Second, data, and I'm sure this one won't get fixed. TikTok cut the search function out of its own Creative Center in 2024 after researchers used it to study the platform. The one public trend tool, and they pulled it. I used to work at TikTok. I know what's on the priority list. The API platform has never been near the top, and for most Chinese companies it never will be. Every analysis product downstream of it is built on sand.
So the work is in the details. Hundreds of small decisions nobody sees. Which is what I told him: don't go big this time. Find one wedge so small it looks embarrassing, and go deeper on it than anyone has bothered to.
And if you actually do that work, here's the trap.
You spend weeks on details nobody sees.
When you finish, you have to prove you beat a user who just prompted a general agent. The same user who, like me, is quietly unhappy with their own output but will never admit it on a sales call.
Then a new model drops, and half your skills, structure and fine-tuning look pointless.
Then you launch. Into a feed where "built with AI in a day" products are everywhere. I don't even open them anymore. One glance, swipe. Nine out of ten are slop, so the reflex is right. But the reflex has a cost. The products built with real care get swiped past too. Not because they're bad. Because they're buried under more noise than the internet has ever had, and I'm one of the people scrolling.
I know founders who built genuinely good things in these categories this year. Almost none of them got found. They didn't fail at building. They failed at being seen.
That's the one thing I don't see getting fixed: distribution. For Chinese teams going global especially, it's not something we're born with. We have to earn it, and most of us haven't yet.
So this is the market right now.
Users with a real need who can't find a product, so they prompt, and aren't satisfied, and tune, and tune again.
Builders who did the hard work, invisible.
And a whole set of categories everyone wrote off as "just a prompt" that are, in fact, wide open.
This is the best era for people like us. It's also the worst.
The hard part was never the idea.
every ai startup in SF is stuck in the same prisoner's dilemma.
we have endless tools and skills that generate videos for us, and the internet has never been worse
those tools were supposed to amplify our taste. instead they just amplified our output
skip the slop and the startup next door posts it and takes your traction. so everyone ships the same AI video with a different logo
@mitchellh is right that it's tacky. but taste is a luxury for companies that already won. for the rest of us, taste is how you die quietly
we're not victims of the slop. we're mass producing it
I have one rule now: no one-line prompts
every day I spend about 30 minutes collecting things I find beautiful, and that's what I feed the model.
i honestly don't get why people brag about making a whole video from one sentence. that's not a flex, that's slop
AI is putting me in a filter bubble worse than tiktok
every day I have codex run research for me. competitors, new launches, what's working. I'm used to it now and honestly I love it.
but something's been bugging me lately.
claude will tell me some direction is really working. I push back a little, and suddenly it's like, you're right, actually this direction isn't worth following.
so which one is it lol
tiktok at least wanted something from you, your attention. claude just wants you to stop frowning. it's not doing research, it's doing customer service on your ego.
half of @vineerpasam's thread is roasting chatgpt for telling people what they want to hear. claude does the exact same thing, it just has better manners about it.
and every founder in sf is running the same agents on the same competitors. that's not research, it's one shared bubble with a nicer UI.
the scary part is I caught myself unable to form an opinion on a competitor without asking claude first. I hadn't just outsourced the research, I'd outsourced the thinking.
so I'm doing something that sounds insane for an AI founder. at least 3 days a week, no AI.
if your best ideas only show up after you open claude, they're not your ideas