How to Write: Advice from David Perell and Sahil Lavingia

[ writing ] · 4 min read

Writing is a superpower. Telepathy to be exact.

Think about it—through writing, I broadcast ideas from my mind to yours. Across time. Across space. (Yes, the internet plays a big role, but let’s focus on writing.) The more effective your writing, the stronger your telepathic ability.

How to write

Why write about writing on this site?

Writing is essential for effective data science. Good writing means you get buy-in on ideas, your methodology and experiments can be replicated, and readers understand enough to give feedback. Poor writing gets you zilch (and snores). Business folk have enough trouble understanding data geeks as it is—don’t make it harder with your writing.

Writing is an important way to learn. (The other important way is learning.) When writing, you have to organize ideas and prune the unnecessary. Along the way, you find gaps in your understanding, which leads you to more research and learning.

Writing becomes more important as your career progresses. Everything is writing—emails, specs, documentation, articles. Code too. As you become more senior, you write more docs, less code. Seniors contribute by designing systems and communicating them to teams to implement.

I recently sat in the excellent workshop by David Perell (The Writing Guy) and Sahil Lavingia (Founder and CEO of Gumroad). Here’s a summary of what I learnt and some of my thoughts mixed in.

How to write

Write what excites you. If you’re not excited about what you write, it’s unlikely to excite your audience. Writing is difficult; thus, just start with something that excites you. Then, build a habit by writing more about what excites you.

Write simply. Communicate strong ideas clearly. Sometimes, people use “clever” writing to hide weak ideas. Think of those convoluted investment prospectuses. And fraudulent company annual reports (read: WeWork). Keep your writing simple to let your ideas shine.

Write before you’re an expert. People think they have to be an expert before writing—why else would others be interested in what you have to say? Answer: Because you have unique experiences that others want to learn about. Interviewed at multiple companies and got an offer? Teach us. Set up a productive dev environment or workflow? Demo it. Expertise is a ladder. Wherever you are on it, there’ll be people above (whom you view as experts) and people below (who view you as an expert). People below want to learn from you.

Write for one person. Make your message focused. It’s better to have a small number of people who passionately care about what you share, than to have a large number of barely interested people. Though you only reach a small percentage (1% of 1,000), over time, that 1,000 grows to 100k, and you reach 1k people.

Write for everyone and you write for no one. Write for one person and you write for thousands.

Write protein bars, not empty-calorie snacks. Make your writing dense and full of nutrition. Assume your audience is smart and busy. They’re likely high achievers who exercise, eat well, work hard. They appreciate high-value, concise content. Sometimes, you see content on social media (I’m looking at you LinkedIn) that’s light, breezy, and has a lot of likes. But when you reread it, you realise… it doesn’t give a lot of value. And you didn’t learn anything from it. Don’t write like that (unless your aim is getting likes).

Write iteratively. Start by pouring all your ideas (don’t judge them, yet) into a draft. Then, rewrite that draft for organization, clarity, and conciseness. Send it out for feedback. Repeat. Sahil shared that he shared his Medium post with 40+ people for feedback, and probably took him 40-50 hours. What you see as beautiful writing is the outcome of many rounds of rewriting.

“There’s no such thing as good writing, only good rewriting” - Robert Graves

Write evergreen content. Focus on topics that will always add value and be helpful. Perhaps a short essay on why and how to write. Such writing stays relevant for years. (Nonetheless, many people write such pieces, so you might not contribute much. But hey, you learn a lot by writing about it). So reconsider writing that 183,768th piece of COVID-19 visualisation/dashboarding.

If you found this post useful, share this viral tweet with your friends. Spread the word on writing effectively. =)

Thanks to Yang Xinyi and Marianne Tan for reading drafts of this.

If you found this useful, please cite this write-up as:

Yan, Ziyou. (May 2020). How to Write: Advice from David Perell and Sahil Lavingia.


  title   = {How to Write: Advice from David Perell and Sahil Lavingia},
  author  = {Yan, Ziyou},
  journal = {},
  year    = {2020},
  month   = {May},
  url     = {}

Share on:

Browse related tags: [ writing ]

Join 6,700+ readers getting updates on machine learning, RecSys, LLMs, and engineering.