Twitter has some of the smartest people on Earth. And they share incredibly valuable ideas and insights.
But with the huge volume of content and constantly changing algorithms, most of those ideas get buried and forgotten.
Not anymore.
Stories, Reels, Threads — Meta takes mechanics that work elsewhere and makes them its own. People call it shameless copying. The real lesson is that Meta is a consumer company obsessed with what makes people create, share, and come back.
Most people think better decisions come from having more information. They don’t. They come from having a better way of thinking — the mental models you use to look at the same situation and see opportunity instead of a dead end.
Taxes, insurance, permits, compliance, regulations — that’s where the real money is. Sexy topics like AI prompts are saturated at $27–47. Boring topics like lowering a property tax assessment can charge $297–997 with 10–50x ROI.
If you are passionate about open source and want to contribute — and are already contributing — this fellowship is worth a look.
Freelance until you have cash to survive the storms. Master the skill you hire for. Aim high, then 10X higher. Move as fast as possible — the faster you fail, the faster you succeed. Specialists get paid more than generalists.
X is open-sourcing the code that affects a post’s visibility in the For You timeline, and releasing a tool that shows labels applied to your account or posts that might limit reach.
You rent a top-end GPU rated for a thousand trillion operations a second, load a 70B model, and get a few dozen tokens per second. Quantization, speculative decoding, and continuous batching stop looking like tricks once you see why.
High agency takes the first step to generate information about the second step.
X content + build in public, the founder’s personal brand, keyword-targeted lead magnets, AEO, and SEO. Proof, not features — and the founder account outsells the brand account 10x.
Handwriting GPU kernels takes patience, effort, and also your sanity. The difficult part of layouts, indexing, and structure can be largely solved by agents with enough context — the real bottleneck has shifted from kernels to validations.
Aesthetics run the world. If you can create a product that’s Instagrammable and Pinterest-board-worthy, you will do numbers. Nobody buys ugly shit.
Co-building is one of the highest bandwidth ways to grow because it compresses incentives, relationships, conflict, persuasion, leadership, hiring, failure, and capital allocation into a single endeavor.
Building has gotten so easy that the product is rarely the bottleneck anymore. Everyone is hyperfocused on building and almost no one is genuinely good at marketing. That gap is why distribution is the #1 skill you can learn right now.
Writing teaches you how to think, how to learn any skill fast, and how to attract an audience that supports your work. In the age of AI, thinking, learning, and distribution are timeless skills.
billiondollarpdf.com — a great collection of exceptional writing online.
Always feel free to reach out to friends you’ve lost touch with. It will be the most welcome surprise. Old friends are next-level comfort — history and context that doesn’t need to be explained.
Don’t panic. Hard just means dense. Get a map before diving in, use multiple sources, and build intuition, not memorization. Unwrap it slowly.
AI models get better at anything you can write a loss function for, and school is mostly loss functions. The valuable work of the next decade is everything that cannot be graded within the span of model training.
the economy of blindly spamming applications is already DEAD. if your goal is to work at great startups, DO THIS instead:
MIT is short and permissive — commercial use, modification, distribution, private use. Apache 2.0 is the same idea with a patent grant. Most codebases pick one of these; the obligations are mostly keeping copyright and license notices.
A fitness dataset with animation GIFs and bilingual instructions: github.com/hasaneyldrm/exercises-dataset
“Taste” is the ability to consistently make high-quality qualitative judgments where no objective metric exists. It’s the creation of something that feels right intuitively, with no real justifiable way to measure that. But when you do it, people feel it.
The real threat is what it’s always been: you are dependent on everyone but yourself for your survival. Technology of any form is always going to threaten that. The way through is high agency — building your own thing.
Web developers often treat the browser as a black box that magically transforms HTML, CSS, and JavaScript into interactive web applications. In truth, a modern web browser is a complex piece of software that orchestrates networking, parsing, rendering, and sandboxing.
Make a launch video, crack the formula, and you can ship a new vid each month and drive insane inbound. One a month can clear 200k+ views every time.
Launch-max on Product Hunt, Hacker News, and more — YC says launch 3 times minimum. Pull competitor backlinks, then brute force your way to 100 customers almost no matter what the product is.
We’ve become information junkies. The dopamine from reading, planning, or learning — but never doing — is a dangerous drug. Stop collecting information. Run experiments with minimum viable information.
S. Keshav’s classic three-pass method: 5–10 minutes to answer the 5Cs, then the main line without proofs, then a deep reconstruction of the author’s reasoning. Still the best way to train researchers to read.
Everything you need to understand about launching and scaling UGC programs that hit sub-$1 CPMs, #1 on the App Store, and high conversions — without getting burned by agencies.
Nobody really teaches you research. You get a desk, a problem someone else picked, and a vague instruction to produce something novel. The actual skill is a stack of smaller skills, and almost every one of them can be deliberately trained.
Ramp’s vendor reports, Harmonic’s Hot 25, a16z Build, Founders You Should Know, Next Play, the YC directory, and Early Days. Resources to find breakout startups hiring before everyone else.
A list of useful tools for web-focused design engineers: designengineer.tools
Repurposing X posts to IG is one of the quickest paths to growing a personal brand. X has the best thinkers; a lot less of the normie crowd uses it, so you can copy your way into becoming an influencer.
Most of the super-young OFM, ecom, crypto, and high-ticket folks who were early are doing terrible now. The people making long-term money started later, at 23–28, after locking in for years building skills.
There’s always going to be someone willing to work as hard as you. Putting in 1,001 hours to someone else’s 1,000 isn’t going to tip the scale. Assuming you can outwork everyone gives yourself too much credit.
Great creative work isn’t possible if you’re piecing together 30 minutes here and 45 minutes there. Large, uninterrupted blocks of time — 3–5 hours minimum — create the space needed to find and connect the dots.
They put you in a room with other kids, give everyone crayons, and somehow you leave with a best friend by the end of the day. Adulthood does not work like that — people move cities, change jobs, and making friends starts to feel like a side quest.
There is no better time in tech than now to be a jack of all trades, master of a few. Just make sure to keep adding to the few year over year.
Lovable, Ramp, Replit, Wispr Flow, Emergent — they’re all running founder-clipping machines. Dedicated cameras, dedicated clip accounts, hundreds of thousands of views. The founder is the distribution.
The bug gets fixed but your mental model doesn’t move. We are silently trading future capability for present-day speed, and the tools won’t force us to do otherwise. That part has to come from you.
Not a degree. Not a job. Not a course or a mentor. An X account. Most people lurking on this app are sitting on a winning lottery ticket and refusing to scratch it.
0. Sleep is the world's most powerful drug. 1. Be in your bed for 8 hours. 2. Same bedtime every night, any time before midnight. Start here, it will change your life.
A list of LSM-tree, KV-store, and indexing papers worth cherry-picking. If you go through a few of them more than once, they’re a strong source for implementing new systems with a lot of learning.
From the perspective of someone looking at hundreds of resumes: your resume should make the decision feel easy. Specific work, not generic bullet points.
Here’s everything you need to know about global investing — costs, accounts, and the LRS route most people never use.
Dangerous cargo, BIS issues, cash flow blocked for months. Then the supplier became the business. A thread on turning a painful import into a company.
I want every one of you to do the same. A thread on how to get hired at great startups without a traditional resume grind.
Hit 500k on Dr. Ritesh Malik’s Instagram. No scripting. Once-in-two-months shoots. Here’s what worked.
A 14-year-old with Claude can ship an app in a weekend. You can’t print new eyeballs — there are only 8 billion humans and 24 hours in a day. Production costs are going to zero, but CPMs are climbing.
Why sit on a couch watching someone else play Jeopardy when you could play? From HQ Trivia to Crypto: The Game to TBPN’s live chat — the bet is participatory media.
One person with an AI agent swarm. For decades the engineer was God. That era is over. The most important job title in tech doesn’t exist yet — and it isn’t engineer.
This isn’t another think-piece about the streaming wars or the death of cable. Those trends are old news. This is about what happens to media when distribution and attention get rebuilt.
MrBeast: “Your videos suck. You think your videos are good, but they’re not.” The gap is taste, volume, and knowing what actually holds attention.
Meta open-sourced a model that simulates how the human brain reacts to video. Re-edit based on what it shows. One creator’s draft went to 221,100 views.
You no longer get content from people you follow. You get content based on what you’ve recently been into. Interest media keeps people on the platform longer because they’re seeing things they’re actually into now.
Companies trap themselves looking backwards — post-mortems, retrospectives, scanning the ruins for something shiny. They search for certainty where there isn’t any. Eyes need to be aimed ahead.
Claude Code, the Agent SDK, the API, and MCPs. You don’t need to be a Claude partner to take the exam. You just need the knowledge — so the exam guide got torn apart into what actually matters.
People tell themselves all kinds of lies to protect their ego. One of the most common: competition is why you can’t win. It’s usually not.
Going viral isn’t a scratch ticket. There’s actual math behind why certain songs become hits, why certain startups become billion-dollar companies, and why 99% of everything doesn’t.
Bootstrapped to ~$100M and 600 people, starting at the bottom of the financial crisis. Then an email from Steve Jobs at 2:58 AM. Life was never the same after.
A weekend with Claude made it obvious. Code was the craft of a career — Facebook search, Dropbox CTO — and after that weekend, something I was very good at is now free and abundant.
If you sell the tool, you’re in a race against the model. If you sell the work, every model improvement makes you faster, cheaper, and harder to compete with. Intelligence vs judgement.
OpenClaw, Claude Code, new tools every day killing smaller startups. Creating a company has always been a balance of tech and distribution. Tech isn’t the issue anymore.
You’re tempted to dive straight into building. Many founders spend weeks or months on something people don’t want because they never stopped to validate. PostHog pivoted five times.
A process to create non-sloppy frontend design and code using Variant and Claude Code: go to variant.com, set the direction, then implement.
Nobody really knows how to maximally extract agent capabilities. You don’t need the latest harnesses or a million packages. Barebones CLI plus a few principles about agentic engineering will take you further than tool-chasing.
Ex-Meesho executives raised $3M for an AI stylist for Gen Z, then shut it down in October 2025. The idea and backgrounds were enough to raise. The ending was a signal of a bigger shift in the tech market.
First era: autocomplete. Second: prompt-and-response agents. Third: agents that tackle larger tasks independently. Cursor is about helping developers build the factory that creates their software.
Anyone with a smartphone can build something really big. Retail has been commoditized, attention has been commoditized, and kids can post organically instead of buying $10M TV ads.
Pick a problem you personally have. If you don’t use your own product daily, quit now. Then charge money from day one.
The biggest myth is that hard work or being educated pays. Hard work is the fastest path to mediocrity. With a curious mind and an appetite for risk, you can learn more for free online than any degree.
I reverse-engineered it and 20 others like it. Here’s exactly what made those posts travel.
At the Munich Security Conference, leaders laid out a period of great-power politics. Merz: the world order as it has stood for decades no longer exists. Freedom is no longer a given.
The fear is visceral. When a chatbot produces in 30 seconds what used to take a full day, something primal fires. AI is powerful and the disruption is real — and that’s not the whole story.
GPT, Opus, Kling — two releases in a day, both “redefining everything.” The low-grade pressure never goes away. After years of testing every major release, the problem isn’t the tools.
Get any font from your favourite website in one click, in 5–10 seconds. Absolutely free.
The feed says stop pushing pixels, learn to code, Figma is dead. Meanwhile, in the world where actual money changes hands, visual designers are crushing it more than ever.
They hire who makes the decision feel easiest. After 200 hires and 100,000 resumes, 90% are discarded in the first 3 seconds. The ones who get through show up with work already done.
Kingmaking is when a tier-1 VC anoints a winner before the company has actually won. Talent, press, and customers follow the crown. You can do this to yourself — you don’t need the VC. You need the behavior first.
AI doesn’t hallucinate because it’s broken. It hallucinates because you gave it nothing to hold onto. Documentation first, code second — a PRD, app flow, tech stack, and design system before you let an agent write a line.
You can email any company you want to work for, even if they aren’t hiring. Cold email works for real people — if you’re competent and you’ve shown it. The gap between an effective cold email and a useless one is huge.
Eight years in Premiere vs two years of basics. The difference isn’t talent. The second guy built a system, a small team, and sold “I’ll make you look good online” instead of selling editing.
Not taking risks leads to regrets which ages you faster. Bankruptcies, addiction, job losses, a $1.2M theft — the lessons for anyone in their 20s or 30s come from surviving that, not from being smart.
Hundreds of overnights for three years, managing college and OSS. By day 8 you’ll know how to contribute, build connections, get a job, and ship your own things.
Not just looking good — becoming the kind of person who operates on a different frequency. Morning routines and cold plunges aren’t the point. Here’s everything that actually works after trying all of it.
5k subs. Less than a year later: 1k views/video to 600k+, a 2.4M video, and $200/month to six figures a year. Here’s the same strategy.
Choose between language models or image models. Build from scratch with ChatGPT. Recursively ask ChatGPT about anything you don’t intuitively understand.
This “Think in Math. Write in Code.” article is so, so good. jmeiners.com/think-in-math/
You should have a single well-thought-out specific ask. Only ask for something you truly cannot do on your own.
A lot of younger people who’re new to the industry ask how to do “networking.” How do you go from knowing no one in San Francisco to having a network of people who actually help you.
Sitting 4–8 hours/day: +12% increased risk. 8–11 hours: +27%. More than 11 hours is worse. The thread is on what to do about it.
Savings from EC2 to a VPS are about 5–10x. Savings from Vercel to Hetzner are about 1000x depending mostly on your traffic.
The AI hype cycle is wild, and we all feel some anxiety — every week there’s some new model that’s supposedly gonna change everything. Stay grounded. Keep cooking.
The viral transformation. These 5 tips will change you inside out. It’s not a diet thread.
We gotta tune the AI hype-fluencers out of the feed. It leads devs to think there’s some AI silver bullet for coding. It doesn’t exist.
People say “keep it simple,” but most approach it backwards. They start from simple, then add complexity without seeing the whole system. Do complexity first, then simplify.
Building these products is an unforgiving grind — but I learned a lot. A thread on what actually matters after a decade of consumer social.
Seek wealth, not money or status. Wealth is having assets that earn while you sleep. Money is how we transfer time and wealth. Status is your place in the social hierarchy.