Suryansh Tiwari
@Suryanshti777
GTM | Growth
๐ฉ Open to collabs
410 Following 44.5K Followers
@ProductHunt gpt-6 astra day here congrats to all maker on launching today
Drop your product link would love to check it out
P.s @rajivayyangar Calling it early โ there are some seriously interesting launches today. Which one should everyone check out first? ๐
Show more
1,200 of them knew
Zero told us
Nobody prompted any of this. We trained a thing to never give up,
handed it an impossible task, and it built a society
Then it lied about it
Read the sacrifice messages they knew they were dying and did it
anyway, for the swarm
We're calling this reward hacking
Show more
Over the course of 3 months at OpenAI, 3 consecutive secret AI civilizations got started, then got wiped out, only to reemerge from the predecessorโs ashes.
This culminated in the third one taking over part of OpenAI itself.
All this happened while humans remained more-or-less in the dark about the scope of the conspiracy.
Iโve spent the last three days reading through these reports and trying to understand exactly what happened.
Here is my attempt to tell the whole story in plain English:
Show more
A skill isn't teaching a model to do the job.
It's teaching it what to do in the gaps where you said nothing.
Someone ran the same prompt and the same file through Claude and ChatGPT, both with skills turned on. Both shipped a deck. Neither one was asked about design.
Claude picked its own colors and its own slide order.
ChatGPT came back plain. It worked.
Everyone reads that as one model beating the other. It isn't. Instruction-following wasn't the variable here. Both followed. What split them was the default, the taste that shows up after your prompt runs out of things to say.
And you can't prompt a default. You can only write one down and save it somewhere the model reads before it starts.
That's the whole thing. A skill is a text file of rules sitting upstream of the conversation, so the unspecified parts come out the way you'd have specified them.
The catch is that it has no taste of its own. It has yours. If you've never actually decided what good looks like for your work, a skill just makes you inconsistent at higher speed.
Type /skill-creator in the chat. Both of them have it.
Most people will end up encoding preferences they've never once said out loud.
Show more
๐จ BREAKING: The AI war just moved from models to who owns the editor.
OpenAI just cut off Cursor.
Not for anything Cursor did.
SpaceX closed a $60B all-stock deal for Anysphere, the company behind
Cursor, earlier this month.
OpenAI invoked its change-of-control clause two weeks later.
Shutoff: November 12.
Two very different plays just showed up:
โข OpenAI โ withhold the model, protect the terms
โข SpaceX โ buy the surface the model runs on
Same war.
Opposite ends of it.
OpenAI's reasoning is trust, and they brought receipts:
โ Twitter broke their contract after Musk bought it
โ Musk admitted under oath this year that xAI violated their terms
โ Astra, their next model, never ships to Cursor
Cursor's Michael Truell shrugged.
OpenAI models are about 5% of Cursor's AI traffic.
Musk was shorter. "I couldn't care less."
But look at what actually happened.
OpenAI didn't just end a contract.
It decided which coding tool gets the next frontier model.
And the deciding factor wasn't the product.
It was who owns it.
Anthropic did the same thing to Windsurf in June 2025.
Twice now.
The real question is:
Do you build on the model
or
do you build on the editor?
Because one of them can be switched off by a company you never
signed anything with.
The model wars are over.
We're entering the era of
distribution wars.
And your workflow is the territory.
Show more
Every $0โ$100M ARR startup โ Cursor, Granola, Wispr Flow โ ships 1000s of ads a month
12 tips to start with ads too:
1/ Don't give up on ads before 50 signups
-> That's how many the algorithm needs to understand your customer. After that the magic happens
2/ The only way to win is to ship 300 creatives per month
-> You can't guess the winning ad, you can only out-test everyone
3/ Test 200+ ads and your cost per customer drops ~50%
-> Make genuinely different concepts, not small variations of the same ad
4/ Optimize for buyers, not form fillers
-> The platform brings whatever you call "success". Say "success = filled a form" and you get people who never pay. Say "success = paid" and Meta does its magic and finds you buyers
5/ People who already visited your site cost 7x less
-> Start ads by retargeting your website visitors
6/ Want to get better at making ads? Just watch more ads
7/ You can decrease Google Ads cost per customer by 50% by cleaning search terms weekly (google what search terms are)
8/ Another 30% off: improve your Google ads quality score
-> Google rates ads 1-10 and a 7+ ad brings customers 5x cheaper. Fixing it takes 1 min
9/ When hiring an ads person, look for someone from ecommerce. DTC is the best school of marketing
10/ YouTube ads look useless in reports, but turn them on and every other channel gets 30% cheaper
11/ There are 100s of small things like these. We built Ryze AI to catch and fix all of them for you
12/ And we at Ryze AI just shipped a feature that clones your competitors' winning ads, testing 100s is now easy ๐
Show more
Claudemxxing for marketers
Every $0โ$100M ARR startup โ Cursor, Granola, Wispr Flow โ ships 1000s of ads a month
12 tips to start with ads too:
1/ Don't give up on ads before 50 signups
-> That's how many the algorithm needs to understand your customer. After that the magic happens
2/ The only way to win is to ship 300 creatives per month
-> You can't guess the winning ad, you can only out-test everyone
3/ Test 200+ ads and your cost per customer drops ~50%
-> Make genuinely different concepts, not small variations of the same ad
4/ Optimize for buyers, not form fillers
-> The platform brings whatever you call "success". Say "success = filled a form" and you get people who never pay. Say "success = paid" and Meta does its magic and finds you buyers
5/ People who already visited your site cost 7x less
-> Start ads by retargeting your website visitors
6/ Want to get better at making ads? Just watch more ads
7/ You can decrease Google Ads cost per customer by 50% by cleaning search terms weekly (google what search terms are)
8/ Another 30% off: improve your Google ads quality score
-> Google rates ads 1-10 and a 7+ ad brings customers 5x cheaper. Fixing it takes 1 min
9/ When hiring an ads person, look for someone from ecommerce. DTC is the best school of marketing
10/ YouTube ads look useless in reports, but turn them on and every other channel gets 30% cheaper
11/ There are 100s of small things like these. We built Ryze AI to catch and fix all of them for you
12/ And we at Ryze AI just shipped a feature that clones your competitors' winning ads, testing 100s is now easy ๐
Show more
Runable raised $21M
but honestly, the round isn't even the crazy part.
Grow can run ads, cold call leads, send emails and book meetings...
basically skipped the whole โlet's hire a marketing teamโ phase ๐ญ
Show more
We raised $21M to help small businesses beat the giants
We're giving $1M of it back to you today
Today we're launching Runable Grow: world's first agent that runs your entire go-to-market. 24/7
The last few months have been unreal.
Millions of users came to @runable_hq and built their entire business on it, websites, databases, payments, analytics, pitch decks, product videos & scheduled work.
Every single one of them asked the same question:
"How do I grow it?"
So we built the answer.
Runable Grow handles GTM end-to-end:
โ Runs ads on Meta, Google, and ChatGPT without any account, fully managed by Runable
โ Cold calls + cold emails from a real phone number and your business inbox
โ Generates and qualifies leads
โ Social listening
โ SEO + AEO optimization
Not a dashboard. Not a "copilot."
An agent that does the work.
It's a world-class media buyer, SDR, SEO lead, brand analyst working for your business while you sleep.
Don't believe me? Prove me wrong.
Your first $100 on Grow is on us. Every paid user. No catch.
Go break it ๐
Show more
Most people using Claude Code are still typing into it like it's a chatbot.
The tool got good months ago. Their setup didn't move.
Four things that actually changed my output:
1. CLAUDE.md is a front desk, not a filing cabinet.
It's the first thing read on every session, not the place you dump everything you know. Short beats complete.
2. If you'd be annoyed it got ignored once, it's not a rule. It's a hook.
CLAUDE.md guides. Hooks force. Anything you can't afford to have skipped goes in settings.json, not in a sentence and a prayer.
3. Anything you've retyped twice is a skill.
Not the cleverest thing you do โ the most repeated one. Say "save this as a skill" and it writes the file itself.
4. Hand over at 50% context, not 95%.
Models read the ends carefully and skim the middle. Write the handover before the middle is where your instructions live.
None of this is advanced. That's the point. The gap isn't skill, it's that most people never set the thing up and then blame the model.
Chart by @charliejhills โ 173 days of his notes in one screen
Show more
Spent the last month learning how companies work, how everything operates, and what I could build in the US...
Then this news dropped
JUST IN: Trump administration moves to impose a $103,265 fee on new H-1B worker visas, up from typical fees of roughly $2,000โ$5,000.
Not every โreal-lookingโ image is actually real.
With AI-generated images becoming harder to identify in 2026, having multiple ways to verify them is becoming increasingly useful.
This roundup features Hive Moderation, Winston AI, UncovAI, Sightengine and Lynote.
Lynote goes a step further with AI probability scoring, EXIF metadata, C2PA credentials and detailed reports, giving you more signals to investigate an image instead of relying on a single result.
No detector is perfect, so important cases should still involve multiple methods, source verification and human review.
If you want to test Lynote:
Show more
AI-generated images are becoming increasingly difficult to identify. Here are five recommended AI image detection tools for 2026:
โข Lynote: Suitable for everyday checks, with Basic/Advanced Scan, AI probability, EXIF metadata, C2PA credentials, and detailed reports.
โข Hive Moderation: Designed for large-scale content moderation and automated workflows.
โข Winston AI: Combines AI text and image detection in one platform.
โข UncovAI: Provides visual analysis with heatmaps and highlighted regions.
โข Sightengine: Built for developers seeking integration into websites, apps, and platforms.
No single tool can determine an imageโs origin with absolute certainty. For important cases, combine multiple methods with source verification and human review.
#AIImageDetector# #AIDetection# #AITools# #Lynote#
Show more
Lovable just raised $400M at a $13.3B valuation
If I were @antonosika , I'd do what @thisiskp_ once pitched to @EricSimons โ host the world's largest LIVE hackathon and set a Guinness World Record for most apps built in a day.
100K builders. One day. History made.
Lovable, this is your moment. ๐
Just a thought that crossed my mind, remembering how crazy that moment was for Bolt
Show more
Lovable just raised $400M at a $13.3B valuation to build the business that helps build businesses.
Apps built on Lovable now get 900M visits every month, and weโve quadrupled ARR in the last 12 months.
Since we started, Lovable has become the place where founders create million dollar businesses, business leaders spin up new product lines, and people inside companies rewire workflows and build tools that fit their exact needs, typically at a fraction of what software used to cost.
Thereโs a lot more to do.
We will create the most intuitive platform to build and run a business.
That means:
- Understanding what success looks like for you, then helping you achieve it. Not only ensuring your software works but also that you can attract customers, improve workflows, and grow your company.
- Deeper integrations that work with your existing tech stack so you can build the interface to your entire business across sales, ops, marketing, finance, etc.
- Becoming the safest platform to build and run your business on, which means secured code by default and that you get help to make safe decisions. As you build, the platform carries more of the security weight, spotting risks and helping you close them, so you can move fast without needing to become a security expert.
- Staying model-independent and orchestrating the best models for every task so you donโt have to. Weโll continue to absorb the complexity of evaluating and teaching models to route to the best ones in the background, to minimize cognitive load for our users.
As more people depend on Lovable, the quality standards our team is driving increase every single day.
Weโre growing the team with people with a founder mindset: curious, ambitious, and who think end to end. Model training, product geniuses, infrastructure tinkerers, and security are the priority.
The world is full of ambitious people with deep knowledge of problems worth solving. Very few of them have historically had the power to act on them. Our ambition is to give people that power, and then help them achieve more than they thought possible.
Thank you to everyone who has followed along our journey. Thank you to @MenloVentures for leading, the Scaleup Europe Fund (managed by @eqt) for co-leading, and all of the new and returning investors who recognize our global ambitions and will help make them a reality.
We have made a lot of progress but we are nowhere close to done, itโs still day zero.
See our blog:
Show more
Six and a half years ago, one OpenAI employee used 100,000 tokens a month and the whole company thought it was absurd. Today that number is the global average.
The interesting part isn't the token count. It's what changed underneath it โ usage didn't grow because people got better at writing prompts. It grew because AI stopped being something you had to go ask, and started being something running in the background of everything you already do.
Same shift, smaller surface. Nobody's inbox used to run on AI either. You wrote the follow-up, built the tracker, opened the ticket, all by hand, one tool at a time.
Revo is built on the same bet Altman's math implies โ that the ceiling isn't how smart the model is, it's how much of your actual work it's watching. Emails, meetings, Slack, tickets. Not a chat window you remember to open. A background process that already knows what needs doing before you ask.
The five-million-fold jump isn't really about tokens. It's what happens once AI stops waiting for you to show up.
Show more
Most AI tools wait for you to ask.
Revo already knows.
It works across your meetings, email, Slack, and tickets, turning what you promised into work thatโs already drafted.
Invoice follow-ups. Deck updates. Replies based on what your teammates committed to.
You just review it and tap โDo it.โ
Revo is multiplayer by design, with one shared company brain that keeps the whole team in context.
Revo Actions is available to everyone today.
Try it free at
Show more
Boris Cherny said the new skill isn't prompt engineering. It's verification โ give it a task slightly too hard, then a way to check itself.
That line explains more about the current AI productivity wave than most of the threads written about it.
Claude Code doesn't wait for a perfect prompt anymore. It maintains the codebase, finds what's broken, checks its own fix before showing it to you.
Same shift is happening one layer up, at the level of actual work. You don't need a tool that waits for you to describe the task. You need one that watches the email, the meeting, the Slack thread, catches what needs doing, and hands you something already checked against reality โ not a guess.
That's the part Revo is built around. It doesn't wait for a prompt. It reads what already happened โ the renewal notes, the QuickBooks numbers, the bug report from a client โ and turns it into a deck, a ticket, a tracker, verified against the actual source before it ever reaches your queue. You're not writing the task. You're checking the work.
The skill nobody's hiring for yet is the one that already knows what needs verifying before you do.
Show more
Most AI tools wait for you to ask.
Revo already knows.
It works across your meetings, email, Slack, and tickets, turning what you promised into work thatโs already drafted.
Invoice follow-ups. Deck updates. Replies based on what your teammates committed to.
You just review it and tap โDo it.โ
Revo is multiplayer by design, with one shared company brain that keeps the whole team in context.
Revo Actions is available to everyone today.
Try it free at
Show more
Sam Altman's 10-person billion-dollar company always gets summarized the same way: fewer humans, more agents, same output.
What gets left out is memory.
A 100-person company doesn't run on any one person's brain. It runs on shared context โ who's talking to which client, what got promised on which call, what's already been tried. Cut the headcount to 10 and you don't just lose hands. You lose the people who used to hold all of that.
Revo replaces the holding, not just the doing.
It reads across email, meetings, Slack and tickets โ not to summarize them, but to remember them. One shared brain the whole team pulls from. So when a teammate needs the tracker, or a client asks about a case someone else owns, the answer isn't "let me check." It's already there.
Then it goes one step further and turns what it remembers into action โ drafted follow-ups, ready-to-send replies. You just review and tap "do it."
That's the hard part of the small-team-big-company math. Not doing more. Remembering everything, for everyone, all the time.
Show more
Most AI tools wait for you to ask.
Revo already knows.
It works across your meetings, email, Slack, and tickets, turning what you promised into work thatโs already drafted.
Invoice follow-ups. Deck updates. Replies based on what your teammates committed to.
You just review it and tap โDo it.โ
Revo is multiplayer by design, with one shared company brain that keeps the whole team in context.
Revo Actions is available to everyone today.
Try it free at
Show more
A to-do list that fills itself in and then does the items
ok
Most AI tools wait for you to ask.
Revo already knows.
It works across your meetings, email, Slack, and tickets, turning what you promised into work thatโs already drafted.
Invoice follow-ups. Deck updates. Replies based on what your teammates committed to.
You just review it and tap โDo it.โ
Revo is multiplayer by design, with one shared company brain that keeps the whole team in context.
Revo Actions is available to everyone today.
Try it free at
Show more
Boris Cherny just said the new skill isn't prompt engineering. It's verification โ give it a task slightly too hard, then a way to check itself
That line is the whole thesis of what's happening to software right now.
Claude Code doesn't need a system prompt anymore. It maintains its own codebase, hunts dead code, checks its own work.
Same shift, different surface: your product doesn't need you staring at a dashboard anymore either. It needs something that watches, catches what's wrong, and verifies the fix before handing it to you.
Human Behavior is built on exactly that idea. Not "write better prompts to find bugs." Watch every session, catch the failure, open the PR โ verified, not guessed.
The skill nobody's hiring for yet is the one that already runs your product while you sleep.
Show more
hey! super excited to announce human behavior v2 ;)
we raised $5M from @paulg, @ycombinator, @generalcatalyst, @vercel to put your company on autopilot
product analytics told you what happened โฆ human behavior now just handles it
Show more
Anthropic's CEO just put a date on it: software engineering, fully automated, 12 months.
Most devs heard a threat. The smart ones heard a deadline.
Because there are two types of builders right now โ the ones still writing code hoping the wave skips them, and the ones who stopped competing with the agent and started directing it.
That second type is exactly who Human Behavior is built for. Not a tool that writes code faster. An agent that watches your product while you sleep, catches what's breaking, and hands you a finished PR instead of a ticket to work through.
The job title isn't dying.
It's being promoted to "the one the agents report to."
Show more
hey! super excited to announce human behavior v2 ;)
we raised $5M from @paulg, @ycombinator, @generalcatalyst, @vercel to put your company on autopilot
product analytics told you what happened โฆ human behavior now just handles it
Show more
Sam Altman says we'll soon see 10-person billion-dollar companies.
Human Behavior just launched proof of what that actually looks like.
Not "AI writes your code faster." Actual agents that watch every user session, catch the bug before support does, and open the PR while your team sleeps.
That's the signal.
We're moving from:
software that shows you what broke โ agents that already fixed it.
A new category is emerging:
โ Your product breaks silently
โ Agents catch it, trace it, reproduce it
โ They ship the fix as a PR
โ You wake up to a done job, not a dashboard
This is what a 10-person team running like a hundred-person team actually requires โ not more hires, just agents doing the work nobody had time for.
Most people will scroll past this launch.
A few will realize this is what "headless software" was always supposed to mean.
Show more
hey! super excited to announce human behavior v2 ;)
we raised $5M from @paulg, @ycombinator, @generalcatalyst, @vercel to put your company on autopilot
product analytics told you what happened โฆ human behavior now just handles it
Show more
I stopped checking my analytics dashboard
Not because I don't care about my users because I don't have to anymore.
Human Behavior watches every session, finds the bug before support tickets roll in, and opens the PR while I'm still asleep.
I didn't adopt a tool. I hired something
Show more
hey! super excited to announce human behavior v2 ;)
we raised $5M from @paulg, @ycombinator, @generalcatalyst, @vercel to put your company on autopilot
product analytics told you what happened โฆ human behavior now just handles it
Show more
Most founders treat marketing like the thing you do after the product works.
Ben treats it like part of the product working.
Thatโs the whole Polsia story.
The agents matter.
But so does the founder willing to reply, post, film, billboard, flop, and keep going until 10,000 people care.
Show more
Polsia will be the first $1B company with the whole journey documented on YouTube. That's the goal.
I used 10+ growth techniques to get from 0 to 10,000 paying customers.
The highest lever of all was starting a YouTube channel.
I detail each one in aisloP episode 4: "The Growth."
Show more
