if i knew these lessons 5 years ago, i'd be up $10M.
sharing so you don't have to learn the hard way.
1. who you sell to > what you sell
in high school, i made $30/hr tutoring other students. felt like good money at the time.
but when i graduated, i realized i had been selling to the wrong person. kids don’t make buying decisions. parents do.
1/ cont
so i shifted the entire pitch.
i rebranded. not just a tutor. a complete college admissions consultant.
i’d already helped friends and family get into top schools. i had results. i also had the pedigree as i went to uc berkeley.
same knowledge, same skills, but now i was selling to bay area tiger parents with $1M+ in income and high expectations.
1/ cont
now i was charging $200/hr instead of $30.
the offer didn't change much, as we still did a TON of subject matter tutoring - what did change was who it was packaged and sold to.
you can be brilliant, but if you’re selling to broke or indifferent people, you’re playing the game on hard mode.
the right offer to the right buyer makes all the difference. i refuse to sell to people without money now.
2. sales is king
i was lucky - i did debate in school, so i already knew how to frame arguments, speak confidently, and win people over.
every time i got a parent on the phone, i closed them. didn’t matter what my price was.
2/ cont
i wasn’t lying. i had real knowledge and experience. but the difference was, i could explain it clearly and make people believe it.
sales is just making someone believe that you are capable. an obvious recommendation to you all is actually be capable so that client refers 5 more (which is exactly what happened to me).
i’ve met people way smarter than me who fumble here. they think being right is enough. it’s not.
you need to be able to sell what you know. if you can’t do that, you’ll keep losing to people who can.
3. if something works, triple down
in college, i landed a $100k consulting project with a fortune 500 company.
we were students. that money was life changing. and once we landed it… we relaxed.
3/ cont
we didn’t ask “how do we get 5 more?”
we didn’t productize it, build a pipeline, or strike while the market was hot.
we coasted. and then the moment passed.
2022 hit, corporate budgets dried up, we graduated and started full time jobs, and the angle of “college students helping companies” wasn’t as viable anymore.
3/ cont
if you find something that’s working, push harder instead of easing up. i don't care if you make $1M, that just means you can make 5.
momentum is a gift, markets shift fast, and that thing that’s working today might not work a year from now. this is something i respect about dropshipping heads, they understand this principle better than anyone.
if you have leverage, use it. if you catch a wave, ride it until it breaks or launches you to the moon.
4. only bet on what you actually understand
at the end of the day, any decision you make in the pursuit of money is a bet. even your 9-5, you're betting that the company wont go under, or fire you, or leave you with no exit opportunities.
i work in tech. i follow the space closely. i use these products daily.
so when i bought cloudflare at $40, it wasn’t hype. i had deployed dozens of projects on it and i understood the platform. it touched $198 last week.
same with nvidia. i doubled down when the deepseek debacle brought price down by 50% because i knew what they were building.
4/ cont
on the other hand every time i tried to get cute outside of tech: crypto, random momentum stocks, names i didn’t really understand like DOGE, RKLB, NIKE, LULU, i got smoked.
i lost thousands trying to time the market or make trades without real conviction.
not because i was unlucky but because i had no edge. i didn’t know what i was doing. there's a reason most investors don't beat the market: they don't live and breathe this stuff.
4/ cont
you don’t need to know everything, you just need to know where you have real understanding and bet hard there.
if you don’t understand something deeply, don’t touch it. conviction only matters if it’s earned.
5. invest in your projects
early on, i did everything myself. i built my own website. i wrote all the copy. i even hand-coded my own cold email infra in aws SES just to save $10/month lmao.
i told myself i was being scrappy but in hindsight, i was just being cheap. I could have spent $500 and made it all back with a single client.
5/ cont
now, i pay for leverage.
i’ll hire an expert who’s written $1m+ in copy to write my landing page. because i know that one great page can close clients for years.
i don’t want generic advice from chatgpt. i want the person who’s 10 steps ahead of me. who’s done it at a level i haven’t yet.
if you believe that you're going to sell, you should invest in it. if you don't believe that you're going to sell, then why are you doing this in the first place?
5/ cont
same with tools. if something works and saves me time, i’ll pay for it. no hesitation.
i used to spend weeks building internal tools just to avoid a $20/month subscription. but if i even value my time at just $50/hr, that trade makes no sense.
buy back your time. it’s the highest ROI thing you can do.
tldr
- sell to people who can pay, not just people who need help
- learn to sell; it’s not optional
- don’t let momentum go to waste
- bet hard when you actually know something
- stop being cheap and pay for leverage
took me years and $millions in upside to learn these. hope it saves you some time.
Everyone's talking about AI Agents for Business, but most haven't actually built one, let alone sold it profitably. I've done both multiple times.
Here's the exact playbook I'd follow if starting from zero today - a complete roadmap from learning the basics to landing paying clients. Let me know if you want a YouTube video around it too.
1/ KNOWLEDGE FOUNDATION (Levels 1-2):
First, master the models and their strengths.
- GPT-4o-mini is perfect for low-level intent classification and basic categorization - cheap and fast at around $0.15-0.60 per 1M tokens.
- Gemini excels with large context windows, ideal for long documents.
- Claude is best when output needs to sound human-written: emails, summaries, anything that gets read carefully.
Next, understand automations vs agents. Truth most builders won't admit: You rarely need full agents. 95% of profitable AI tools are automations with smart features. Automations take input, process it, give predictable output. Agents think, adapt, make sequences of decisions - they're complex and expensive.
Default to automation unless you absolutely need the complexity.
2/ TECHNICAL SKILLS (Levels 3-4):
Start with Make or n8n (I prefer n8n for flexibility). These are automation builders, not agent platforms. You'll need to learn webhooks, HTTP requests, API calls, JSON handling, and error management. Find YouTube tutorials for each concept, they're much better than anything I can explain in a tweet.
Build something real that solves a boring problem. Auto-schedule calls, categorize support tickets, whatever. Use it yourself first. This becomes your first portfolio piece and teaches you what actually works vs what sounds good in theory.
Then get technical to scale profitably. AWS is your friend: Lambda for serverless functions, DynamoDB for databases, S3 for storage. Start with AWS's free tier and their own tutorials.
Why go technical? n8n costs $20-50+ per month but is 1000x easier. Going technical drops costs to under $10/month and opens bigger projects with higher margins.
nothing new fixes an old problem. happiness has been a struggle since the dawn of time. it doesn’t come from instagram or SSRIs. it comes from family, friends, and meaning.
depression isn't real because happiness isn't binary; your happiness reflects your inputs. 'depression' is a signal that your inputs must change.
clichés are compressed truth. they’re repeated because they're true, not because they're clever. (i.e. history repeats itself, misery loves company, don't put all your eggs in one basket, ...)
money solves external problems, not internal ones. happiness is an internal problem.
2. intelligence, agency, & outcomes
to be great at something, you must sacrifice almost everything. the greats go mad from being alone with their obsession for too long.
ai commoditizes intelligence. agency is the bottleneck. if you can work relentlessly on your own priorities, this is the easiest era to get rich. the window is closing.
iq and wealth are loosely correlated. the richest people i know aren't geniuses. there's an art to their simplicity i'm learning to respect.
3. skills & habits
everything is sales. every job, every relationship. if you can’t sell, you're capped.
every skill is a muscle. use it and it compounds. ignore it and it withers. you can get better than 99% of people at any skill in 6 months. it just takes focused, hard reps.
your habits become your handcuffs. what feels like a choice is a dependency. your nervous system learns to expect its fix, and becomes irritable without it.
Your vibe-coded SaaS is a security breach waiting to happen.
Cursor and Windsurf will happily ship the leak.
Even @InterviewCoder leaked secrets early into launch.
As someone who has built multiple production-ready applications with thousands of users, from just Cursor with minimum interaction:
here’s a simple system to remove 99% of vulnerabilities, with prompts you can paste straight into Cursor/Windsurf👇
1. Secrets:
Most vibe-coded apps leak secrets by hardcoding API keys. Cursor and Windsurf won’t protect you unless you do it right.
- Put public/non-sensitive defaults in .env (e.g. NEXT_PUBLIC_API_URL=...)
- Put secrets and overrides in .env.local (e.g. SUPABASE_KEY=...)
- Add .env.local in your .gitignore file.
Fix it in Cursor/Windsurf:
“Use process.env. for every secret. If the code needs config, tell me the exact key to add to .env.local. Assume .env.local already exists.”
Then deploy secrets in Vercel → Settings → Environment Variables.
Any host (Cloudflare Pages, Railway) works - just keep your keys off GitHub.
2. Supabase:
Your Supabase anon key can read every row by default. If you skip setup, anyone can open DevTools and dump your entire DB.
Fix it in Cursor/Windsurf:
“Enable Row Level Security on every table. Create policies using auth.uid() so users only access their own rows. Never use service_role in client code. For privileged actions, wrap SQL in RPCs and call them from the server. Always use parameterized queries.”
“ I’m building a [description of your product - the more detailed the better]. Use Next.js for frontend, Supabase for DB + auth.
Give me the full architecture:
- File + folder structure
- What each part does
- Where state lives, how services connect
Format this entire document in markdown.”
Save its output as architecture.md and throw it in an empty folder where your project will live.
“ Using that architecture, write a granular step-by-step plan to build the MVP.
Each task should:
- Be incredibly small + testable
- Have a clear start + end
- Focus on one concern
I’ll be passing this off to an engineering LLM that will be told to complete one task at a time, allowing me to test in between. "
Save it as tasks.md. Again, throw it in the folder.