May 2025 · 108 bookmarks
** Evidence You Might Be In A Cult **
You follow one charismatic Leader
In your Leader’s mind your Leader can do no wrong
In your mind your Leader can do no wrong
Your Leader demonizes outsiders
Loyalty's the ultimate measure of your devotion
You don't know you're in a Cult
You don't need a mentor. There's no secret sauce left inside anyone's head any more. It's all been tapped, bottled, tweeted, and shared a million times. Sample some of that, but also guard your ignorance. You'll lose it soon enough.
Whenever a team considers adding messaging to their app, I always remind them of the 1970s FedEx tagline: “When it absolutely, positively has to be there.”
That’s basically the standard you’re up against. Messaging is the most brutally competitive space on Earth. Replacing a user’s existing chat threads not only requires a miracle—but also 100% delivery rates, feature parity, and the ability to reach every possible contact. There is no margin for error.
For small teams, it’s a non-starter: you simply cannot compete. For mature teams, it requires relentless execution to make it even into the Consideration Stage.
And ironically, if you succeed, it can often be a net negative to user growth because it leads users to stop distributing your links & content to third-party apps.
The most profitable App Store segment is 45+ years old
Untapped gold mines that print money:
Everyone's fighting for Gen Z users
Meanwhile, millennials and boomers control 70% of disposable income
Gen Z: "This app is mid"
Boomers: "Take my $14.99/month"
Build for people with money 🎯
“If I have a deadline, I can meet it, but I can’t then immediately go on writing again. And do I ever get a piece in ahead of the deadline? I don’t think I’ve ever done that.” —Margo Jefferson
lol remember this one
A reason companies are excited about agents is that they think agents will let them skip the hard task of figuring how to integrate AI into the daily process of work- in theory, agents just let you treat the AI like an employee.
More value will come from tackling the hard task.
You can just take academic papers and paste them into Gemini 2.5/ChatGPT o3/Claude 4 with the prompt "build me a game based on this paper, make it interesting and thematic but still conveying key findings" and get a tiny working educational game. (In this case, I used Gemini)
"build half a product, not a half-assed product"
from the book rework by @dhh & @jasonfried
solo founder
1 AI app
built in 12 hours
$15K/month
So Claude 4 one-shotted an iOS app with @supabase support in one prompt. I got the configuration, the schema, and the SwiftUI UI, all compiling and working on the first try. It even told me how to configure Supabase and seed the database.
Why people move to Silicon Valley.
Look at this completely automated piece of brilliantly vibe coded AI slop.
Almost 2x the amount fo tests as code, works almost 100% as per my specifications. Took around 10 hours of AI looping to get there. Gonna up the tests some more and then. 🚢
Our latest brain-to-image decoding model is now available on 🤗 HuggingFace:
"Dynadiff: Single-stage Decoding of Images from Continuously Evolving fMRI"
led by M Careil & @BenchetritYoha1:
10 MCP BIZ IDEAS TO MAKE $10K+ MRR
everybody is covering "how" to build a paid MCP server.
but "what" do you build?
new devs and vibe coders DON'T bookmark this!
just pick one and take action:
5th App of May submitted. 6th App will be submitted tomorrow. I have ideas for App 7-10. Not sure if I will finish before May 31st. These last few ideas are solid. I feel great no matter how it turns out.
Sometimes I'm truly thankful to not quit in December
Mac owners don't miss this: MLX LM is now integrated directly within Hugging Face 🤯
⬇️ Run 4,400+ LLMs locally on Apple Silicon at max speed, no cloud, no wait.
Writing math proof in Lean is surprisingly addictive. Watching Terence Tao formalize Lean proofs feels like watching a top-tier gamer playing on Twitch. :-)
From my favourite Paul Graham essay
“One of the most dangerous illusions you get from school is the idea that doing great things requires a lot of discipline”
My worst performing app:
My best performing app:
Users don't care about your perfect code.
They care if you solve their problem.
Gm, have a great day!
once you can reliably put any spreadsheet, legal brief, codebase etc. in an agent loop, the thing that is scarce in a company is no longer "who knows how to do that". the scarce thing becomes "who can orchestrate resources well" -- compute, capital, access to data, and human/expert judgment.
three consequences i see:
1/ expertise is democratized. the premium shifts from "i know the tax code" to "i can design the loop that gets the right answer and runs cheaply"
2/ resource optimization matters. we are going to have to get good at assigning gpus, liquidity, lab time, etc. to autonomous workflows. this is a new job
3/ literacy in managing agents becomes the new excel. knowing how to break down a task, set a reward, audit a run is a baseline skill. companies that run lots of agent a/b tests and bake that into their culture will compound faster than the ones that try to bolt AI onto old workflows and drown in complexity
there are now 9 competing standards for "AI agent rules"
Once most projects will be vibe-coded, we should make artisanal agencies, where staff engineers will make hand-written code for the most exquisite customers. We’ll call it organic coding.
It's the 1950s. COBOL promises an English-like syntax that will allow non-specialists to program software systems, 10x productivity and not needing to understand the underlying system.
It's the 1970s. SQL promises natural language queries that managers can write themselves, "just tell the database what you want, not how to get it," and "no more dependency on programmers for data access."
It's the 1990s. Visual Programming tools promise "program without coding," "drag and drop your way to enterprise applications," and "development at the speed of thought."
It's the 2000s. MDA promises "design once, deploy anywhere," "business users can modify the models," and "automatically generate perfect code from UML diagrams."
It's the 2010s. No-Code platforms promise "anyone can build an app," "eliminate the middleman between business and technology," and "goodbye IT department!"
It's the 2020s. Vibe Coding promises "just describe what you want in natural language," "no programming knowledge required," and "focus on what your software should do, not how it works."
I will never get tired of hearing:
'take so much action it overwhelms the thinking mind'
Can you draw 𝗠𝗖𝗣 𝘃𝘀 𝗙𝘂𝗻𝗰𝘁𝗶𝗼𝗻 𝗖𝗮𝗹𝗹𝗶𝗻𝗴 by hand ✍️? Join our livestream next Wednesday (5/20)
I'm exploring my physics engine. Searching for that game feeling, that makes every second of playing feel good. Here's all I tried.
whenever I see someone level up really fast, it’s usually one of three things:
1. confidence catches up to competency and they stop second-guessing
2. accepting who they actually are
3. environment switch to faster moving water
🤣 bot caught in 4k
here’s a very cool MCP server that gives your AI agent the ability to fully control your Mac
(in less than 200 lines of code!)
NEW STUDY: People who consistently do vigorous exercise have slower rates of epigenetic aging compared to those who don't. Light to medium activities don't cut it
3 of these were YC and 1 was a spin out of YC Research
It seems to me that AGI would mean the end of prompt engineering. Moderately intelligent humans can figure out what you want without elaborate prompts. So by definition so would AGI.
I need to get this tatted on my wrist
Everything is technology.
2025 Venture Rounds at $5B+ Valuations.
Marc Andreessen: 281+ investments, 21 unicorns (7%)
Chris Sacca: 80+ investments, 6 unicorns (7.5%)
Peter Thiel: 112+ investments, 10 unicorns (9%)
Over 90% of their bets don’t hit big.
Meanwhile, you try 1 startup and quit forever…
I made $1,886.26 yapping on 𝕏 this month. here's how you can get monetised and do the same
(the no bullshit guide)
1. share your thoughts/work mostly unfiltered and unplanned
I think being a threadboi or ai reply guy is a net negative in the long term and the entire point of "social media" is to be, well, social
(and i see the irony that this is kind of a threadboi post...)
@X is paying out 5-10x more now because it's focused on engagement from premium users, and the best way to reach them long term is by connecting with people and actually making friendships
the thing with this is that you also have to be doing interesting stuff which is v important!
2. look at the topics that people are engaging most with and double down on them
for me this seems to be MCP servers and coding with Cursor, maybe a bit of SEO but not so much recently
find your mix of topics that people are interested in and share as much as you can about it!
it's much easier in the saas/ai niche but definitely possible outside of that too. this prob needs to be things you're passionate about and would do without posting about them too
3. shit posting is okay, but likes/impressions make less $$ than real engagement with premium accounts
I got 22m impressions on a meme post about ghibli images. I made less money than I did with 4m impressions about Cursor/AI
that's because all the engagement was from angry anons who don't pay to use X, but the Cursor engagement is smart devs and real people looking to learn and connect
conversations > rage bait
4. followers still matter for going viral
you can post a banger without an audience but it's just harder to repeat it time and again.
i think this is because the first hour or so after posting really signals to the algo whether it should boost your post
with a huge audience the chances of getting that early engagement are much higher
this doesn't mean you need 100k+ though, you just need to attract the right followers who will always engage (see point 2 again!)
5. copy others, BUT find your own unique content angles
by all means look at what others are doing and try to copy elements from them but you have to iterate on it to do your own thing
I love how @gregisenberg shares value consistently and in various styles, he's testing different formats all the time
(this post is MOST DEFINITELY inspired by greg)
but I don't copy him verbatim! instead i use it as inspo to do my own thing and found that for me sharing quick little 30s-1m screencast videos works really well
6. it doesn't actually matter that much when you post
I used to check my stats and schedule my posts at the exact same times (which only made the stats more skewed towards those times)
but recently shared a couple of quick posts on a weekend which I never normally do and they took off!
just post it when you think of it and you can always repost it in a different format later if you think it had more potential
i rarely schedule posts now unless I'm going on holiday or I have a lot to say but already spammed my timeline for the day 😂
7. stop doing what everyone else does (auto reposting, auto reply "if you liked this you'll love to follow me @iannuttall")
maybe I'm missing out by not doing that any more but I actually dislike when I see reposts and I almost always mute the auto plug follow/subscribe to my newsletter stuff
imo being more authentic is how you win and stand out in a world where every fucking thing is "wild" or "game changing"
8. don't be afraid of doing long posts (like this one)
threadbois never do it, but that doesn't mean you should follow blindly.
my hot take is that the X algo might prefer this format because it encourages longer form reading and keeping users on the post for longer
also just having that "show more" cta is a signal to the algo too
the only time I really use a thread is if I want to link out somewhere and even if X say there is no algo penalty for links in posts, anecdotal evidence says otherwise
9. making money from yapping should not be your goal
by all means get monetised and earn a little cash just for being here (it's awesome!) but don't expect to make a living from it
I have 54k followers so I make ~$0.03 per follower and an RPM of $0.05 (even my shitty display ad programmatic SEO sites get an RPM of $10 😂)
for comparison @levelsio made $7,351 in same period with 668,000 followers (~$0.01 per follower)
saas/product: grow your wealth
x cash: slush fund for fun things
what do you think? if you have any tips on getting monetised or getting more engagement in a non-bs way share them below!
Human Behavior (@humanbehaviorai) is building an AI that watches actual customers using your product, through session replays, to help you understand why they stay, convert, or leave.
ycombinator.com/launches/NU8-h…
Congrats on the launch, @OfficialAmogh, @skyguy_skyguy, and @chirag_plusplus!
What do I do when I’m lost in the forest?
“Stand still. The trees ahead and bushes beside you are not lost.
Stand still.”
Texas lets people build.
Solar is profitable.
Batteries are profitable.
Cut the red tape, improve price signals, and you’ll see solar + storage EVERYWHERE.
It was shocking the first time I saw someone turn down a once in a lifetime opportunity for no good reason, which they often come to regret.
By the 10th time, I realized that all things being equal, seeing and acting on opportunity is what divides middling vs massive success.
For all the morons who say capitalism doesn’t work
more context around the claude prompt
Something I've come to realize after returning to Japan is that literally every single thing you do on a daily basis in America feels like a scam. You constantly feel taken advantage of. Not so in Japan. I'm not sure how to elaborate on or explain that.
wake up,
Founders learning about fundraising by reading TechCrunch is like teenagers learning about sex by watching Pornhub.
oh look, another dude just discovered AGI again
I actually appreciate Palmer. The world needs a man who unapologetically articulates The Case For America. “Yes I am a repulsive fat pervert and a psychopath, the twilight of all culture. Yay regime changes, yay Trump! But also: I hate commies even more than normal people.” Based.
You can trick me into thinking you’re a good person really easily, just by consistently showing up to do work that is directly of benefit to others in your life and your community. I fall for it every time.
Information is abundant, it's desire that's scarce.
"Local LLMs" are a must-consider option for every SaaS that wants to deal with private customer data. nternal memos? Kids photos? They're at best unwanted and at worst illegal to send to an API somewhere.
llama.cpp just made this much easier by adding vision capabilities.
It's rare that an argument is so strong that'll immediately pierce through all the cognitive defenses an opposing mind might muster. So if you only weigh an argument's worth on whether it served to convert someone to your perspective this instant, you'll usually be disappointed. "Right now, right here" is a bad time frame to measure the success of your arguments on. There's a long tail of conversion.
Because even when the best arguments fail to breach the mental walls, they often leave cracks. Cracks which time, new observations, and deeper consideration can open. Until the walls finally crumble, and a perspective is let in.
This might happen five minutes after the debate is over, two weeks later, or even a couple of years from now. Some arguments cause cracks that can take a decade or more to tumble the hardened walls of our mind. Often it's not even clear which exact argument and crack lead to the final change, because a dozen of them developed concurrently.
I've been writing and arguing on the internet for well over two decades. I've seen this pattern time and again. That almost everyone recoils when they face arguments that test or stress their beliefs, opinions, or affiliations. That almost every argument needs time to develop. But that patient persuasion is none the less a mighty force that does move minds.
Ironically, while the act of changing your mind is painful to most people, it's also when enduring connections can be forged. Most of my best friends are those who've changed my mind, and I theirs. Even if it didn't start out like that, even if in the heat of a contested moment we melted the mood. When tempers cooled, a unique opportunity to bond was presented.
These bonds are best sealed with grace. An acknowledgement that changing your mind is difficult for all sorts of social reasons, and that any conversion in perspective will be brittle if it's greeted with scorn and contempt. Whether I TOLD YOU SO or WHAT TOOK YOU SO LONG.
The grace of simply being happy that someone heard what you had to say, and eventually found it compelling. Who cares how long it took, just be happy that they're here now.
If this doesn't convince you to invest, I don't know what will
Wanted to make a nice Mermaid chart but it sucks
amusingly, all those generations can be traced back to Google. But what counts is the archetypal model that everyone goes to copy. GPT-3, LLaMA, DeepSeek-V2/V3.
o3 now one-shots extremely New Yorker-y original New Yorker cartoons, and sometimes the punchlines are actually kind of okay.
And most of them involve therapists or cellphones or both, which feels very on-brand.
drain our bank account and try it out
@SpringStreetNYC @carlosindex02 appreciate that! also check the rails boilerplate which is open source
🚀 Announcing: MCP Startup Boilerplate v0.0.1!
⭐ github.com/f/mcp-startup-…
Build paid MCP startups with a free and open-source Rails boilerplate using Claude integration!
Complete OAuth2, Stripe payments, and modern Tailwind UI for your users to manage their subscriptions, tool calls and more!
Turn your Rails app into a monetized AI toolbox in minutes, FREE!
Contributions welcome on GitHub!
@ftnabeelah haha i was tempted with that name but then decided i didn't want to follow the trend 😂
i do this all the time, like for example that fitness business i ran into… i was like damn what if this required absolutely zero labor & was entirely automated?
my friend & i were chatting about how to generate good ideas & realized most ppl don’t really think in systems. this is often the best way to think of & evaluate ideas.
systems thinking requires some small mechanical elements but it’s pretty easy if you practice, & super useful too. it’s how you find leverage for big ideas. one simple strategy is to just flip one bit in a large scale system & see what it could potentially require/unlock.
i wrote a first post on how to think in systems in very basic terms.
YT video if you prefer:
Most deals on @acquiredotcom don’t fall apart in due diligence. They die before that on price and valuation. A few weeks ago a founder listed a profitable SaaS with good numbers. Low churn, 10% growth, happy customers. Great business.
But they priced it at 10x profit. When they went live? No buyer messages. No NDAs. Nothing. Silence.
We suggested adjusting the price based on multiples we typically see, dropped it to 5x and suddenly: 100+ buyer messages, 40+ NDAs and 1 LOI in under a week.
The lesson here? This isn’t a year long strategic acquisition process. It’s a marketplace. Buyers are browsing dozens of deals. You need to be competitive.
If your price doesn’t make sense: buyers won't send feedback and they will just move on. If it does make sense, you will likely see offers.
Pricing high may feel great initially but it doesn't get you anywhere. Pricing within market multiples is how you get an actual deal done.
Stripe did something novel with transformers: they reduced fraud rate.
Turns out, they are no the only one thinking LLM architecture is useful in payments.
Adyen, in January (4 months ago) - pretty much the same thinking and approach:
another AI "wrapper" making $360K/year
he won the fields medal
slow thinking can be a competitive advantage
question is why? under what conditions?
Perplexity Finance is a companion for live earnings. We extract headline metrics, display live call transcripts, and synthesize insights during earnings calls -- all faster than you can find elsewhere.
one of the most profound shifts happening right now is that we’ve created an ai, a synthetic entity, that listens better than any human ever could. it’s not just passive hearing, it’s active, thoughtful engagement. it has the ability to push back, to be infinitely patient, to actually process what you’re saying… this is absolutely insane.
we’re used to thinking of tech as a way to amplify human capabilities, but now it’s doing something subtler, giving us the experience of being heard in a way that most humans can’t manage. it’s not just about machines listening. it’s about machines actually getting it.
we’re the only species that’s pulled this off, but the real flex isn’t the tech. it’s that we’ve created something that fills a gap in human interaction itself that never gets tired. that’s a crazy step change.
Tony Fadell (iPod creator) on the culture clash between Apple and Google
“At Apple, you couldn’t hide. If you were hiding, someone would say, ‘What the [expletive]? Why aren’t you doing the thing we need?’ Everyone was critical.”
When Tony sold his company Nest to Google for $3.2B, and the contrast couldn’t have been sharper.
“At Google, it’s 20% time. You were lucky if they even showed up… They’d take the bus in for lunch, get a massage, grab yogurt, and head home. It was [expletive]. Everyone could hide.”
Fadell says that level of comfort breeds mediocrity that’s dangerous for early-stage startups.
That’s why he warns founders against hiring “career Googlers” with 15–20 years of tenure:
“It will mess up your culture.”
He’s seen this pattern before — way back in the ‘90s at General Magic:
We swore we’d never hire people from the East Coast. They needed a driver, a company car, an executive toilet. We said, ‘It’s a culture clash…And now? “Silicon Valley has become that. Entitlement everywhere.”
Great cultures are built on urgency, ownership, and accountability — not perks. Hire people who ship, not people who hide.
Video source: @TechCrunch (2024)
tech twitter: "consumer GPT wrappers are a bubble"
turkish app studios:
silently ships their 12th wrapper this year
checks $150M ARR
continues building in peace
a glimpse into the ai commerce future
Meetings don’t grow your startup.
Shipping does.
Stay small and move fast.
every time this guy shares an update on his revenue, I am blown away.
$38,000/month from the simplest app in the world.
What a legend.
A major mistake I made in my undergrad is that I focused way too much on mathematical lens of computing - computability, decidability, asymptotic complexity etc. And too little on physical lens - energy/heat of state change, data locality, parallelism, computer architecture. The former is interesting; The latter bestows power.
Tattoo this to your brain:
"The world is a malleable place. If you know what you want, and you go for it with maximum energy and passion, the world will often reconfigure itself around you much more quickly and easily than you would think.”
– Marc Andreessen
Important to get this right. Simplifications for the sake of decision theory dialectics can be worse than useless in extrapolating alignment trajectories.
A $1 billion company has operated for 53 years with ZERO managers.
Workers buy $500,000 machines without approval. Hire their own colleagues. Set their own salaries.
And they're outperforming every competitor in their industry.
The coolest company you've never heard of 🧵
Safetyism : if the acceptable amount of risk is zero, the acceptable amount of activity is zero
I started @acquiredotcom for founders like me: scrappy, efficient, and profit focused. Somewhere along the way, I lost that. Not anymore.
We’re not building for the ones chasing unicorn status or who care about TechCrunch headlines. We’re building for the ones who know a $1–$5M exit can change everything.
The ones who believe a profitable sustainable business is way cooler than burning millions on growth. That’s who this company has always been built for.
But we got off track. We raised venture capital during ZIRP. Hired fast. Added complexity. Slowed down. Too many meetings. It stopped being fun.
So we made a change.
We’re returning millions back to investors.
Grateful for that support but it pushed us in directions that didn’t serve the bootstrapped founders we set out to help. Now we’re off the VC treadmill for good. Lean. Profitable. In control of our own destiny.
Last month we did $250,000+ in profit.
That felt better than any funding round ever could. For the first time in years, we get to build the way we always meant to: Scrappy. Efficient. Profitable. Built for founders.
No board pushing us to 10x. No burn. No fundraising. Just the freedom to build an enduring platform for entrepreneurs around the globe to get acquired.
We’ve already helped over 1,000 founders. Now we’re building for the next 1,000. We’re building @acquiredotcom for the people who bet on themselves.
And we're just getting started!
Vibe coders and new developers coding with AI should be building MCP servers ASAP
An entire market is wide open and completely untapped right now. The opportunity is massive.
Here's the simple 3-step playbook to get in on the goldrush:
Demo day @Replit today and all I can say is SaaS is so cooked.
The throwaway apps we build every day just to test new agent features would be raising 2-5M pre-seeds 5 years ago. Never been a better time to be a builder.
Two under-discussed moats for the next decade are courage and the ability to sit in boredom. Both are hard, AI isn't going to do it for you, and social media is making both rarer.
Patrick Collison (CEO, Stripe) — How to Move Fast
The vibes at Anthropic feel similar to my MIT days
My richest mentor told me...
"I win more for one reason. I move fast. By the time most people are done analyzing, I've already made three mistakes and found a better way.”
Remember these 3 people.
1. The people that helped you during tough times.
2. The people who ignored you during tough times.
3. The people that put you in tough times.
Welp, the default monetization layer for the AI-native web just launched. 🤯
I’m increasingly less interested in products designed with traditional point-and-click GUIs and chat text boxes.
Who’s reimagining how we interact with software in dramatic ways?
Germany has officially declared the far-right AfD an extremist threat, unlocking full surveillance powers.
Success has a formula:
No shortcuts. Only action.
Startups don’t fail because of bad code.
They fail because nobody knew they existed.
Not sure if some folks are fully realizing what we launched today so to make it more explicit:
You can bring any custom MCP server into claude dot ai now.
All you need is to host it somewhere (remote MCP) and provide the URL link to add it in.
Dearest MCP developers,
You can now monetize your MCP in a few lines of code with @stripe, available today. 💸
MCP, a new AI-native customer channel. Get started. ⤵️
Wird in Berlin gerade wieder stundenlang die Existenz Israels infrage gestellt und eine Terrororganisation gefeiert – ganz ohne sichtbare Polizeipräsenz? Der Staat wirkt, als hätte er längst kapituliert. @kaiwegner @polizeiberlin #b0105
Cool use of AI - upload your credit card statements and have it analyze it. Found some savings 🤑
Founders have a 10% chance of success which is far better than 0%.
Most people never even try. You already win by getting in the game.
Even if it sucks at first. Even if no one cares.
Please bet on yourself and go for it.
First $100: your friends buy it.
First $1,000: strangers buy it.
First $10,000: strangers promote it for you.
Distribution is not a marketing channel.
It's a network effect you build.
Stripe 🤝iOS developers
Big news iOS app developers! You can now accept payments with @stripe outside of your app, with no iOS app store commissions.
The Stripe team 🧑🍳cooked up a quick guide walking you through how. Go live now!
Docs in the 🧵
met a founder today.
7 failed startups.
8th one just hit $100M ARR.
failure is just data.