Andrew Hao
banner
andrewhao.com
Andrew Hao
@andrewhao.com
Engineering at YouTube. Previously Lyft, Carbon Five.

Any-surface runner 🏃🏻‍♂️ and your favorite pair partner 🍐. Also: https://mastodon.online/@andrewhao. Account here is no longer me: twitter.com/@andrewhao
Last week was Week 1 of setting Claude Code loose on some Wejoinin errands: having some aha moments about agentic cognitive offloading. Definitely feels like I can accomplish more, but it feels a loose cannon. Bulky, unwieldy, verbose, and sometimes surprisingly effective.

Tons of babysitting.
October 4, 2025 at 5:53 AM
Writing is thinking. I don't think I'm going to pass that off to an AI. I need to work this out by putting words to paper.
July 10, 2025 at 1:03 AM
Oops, forgot to say! This week I imported all my tweets from Twitter over here (thanks @en.blueark.app). Then nuked the Twitter account. I'll be primarily on Bluesky from now on!
February 9, 2025 at 5:19 AM
Scariest night I've had in a very long time, trying to figure out when or where we will have to evacuate. Grateful to firefighters and other first responders for being out there. #eatonfire
January 9, 2025 at 4:34 AM
Taking a solid 2 week break over the holidays was so refreshing. It's been a while since I truly got work off the brain. I felt creative for the first time in a long time!
January 8, 2025 at 1:17 AM
Reposted by Andrew Hao
Non-sequitur "an exploratory project that can't fail" with me the other day. But managing projects in a rapidly expanding product is a whole different deal than managing exploration. Here's how I think about handling the risks of runaway growth: tidyfirst.substack.com/p/managing-e...
Managing Expand Projects
So if “the exploration that cannot fail is not an exploration”, then how are you supposed to manage high-risk, high-consequence projects?
tidyfirst.substack.com
January 2, 2025 at 6:21 PM
Reposted by Andrew Hao
This is (in my opinion) one of the most important pieces of empirical research to come out about generative AI’s actual/likely effects on politics.

@shelbygrossman.bsky.social and team are right to warn that AI built to parrot back our beliefs to us can be an effective vehicle for propaganda.
A new study out of Georgetown and Stanford found that propaganda generated by GPT-3 (not 4!) was persuasive to almost half of the people who read it. I spoke to one of the authors and wrote about the implications www.platformer.news/ai-generated...
February 21, 2024 at 2:39 AM
Reposted by Andrew Hao
my mom named her kid (me) “蓝天” which means bluesky in chinese, so i called myself “jay” to become a bird instead, and then some birdapp company named their project “bluesky” and i ended up running it.
Tell us a true story from your life that sounds made up.
May 1, 2023 at 7:39 AM
Working at Google has taught me a couple of important things, both true:
- The power of written, debated, principled thinking to make guiding product decisions 🏛️
- The demotivating power of a large bureacracy 😭
November 19, 2024 at 7:06 PM
I love data scientists. They are amazing. That is all.
November 19, 2024 at 7:06 PM
Reposted by Andrew Hao
git discussion bingo card
November 23, 2023 at 6:11 PM
Tried Threads and Mastodon. Tryna be active on Bluesky too. Will probably spread some engagement across all three in the next few months - we'll see how this goes!
November 19, 2024 at 7:06 PM
Much love to friends and former coworkers affected by layoffs @lyft. I know some really, really great people and would be happy to help. DMs are open!
February 5, 2025 at 10:07 AM
@mybluewristband shares how she took initiative to shape process where she was at to take control of her own career growth as she grew into eng management. “If I can do it, you can too!” #leaddevsanfrancisco
February 5, 2025 at 10:07 AM
🎨 Slides for my talk are up! It was so fun talking about effective writing with the Pyramid Principle with y'all at #LeadDevSanFrancisco! I'll post some more thoughts later this...
February 5, 2025 at 10:07 AM
Loved @plumertzi’s talk on the UX of incident management at #leaddevsanfrancisco. Moving from a fixed set of heroes to a truly distributed on-call model takes work!
February 5, 2025 at 10:07 AM
Will you be at @TheLeadDev San Francisco next week? I'm excited to be sharing about the One Cool Trick you can use to write Devastatingly Effective Docs™️. It's not too late to sign up - hope to see you there! https://leaddev.com/leaddev-san-francisco
Upcoming events
Join us throughout 2024 for inspiring conferences, worksh...
leaddev.com
February 5, 2025 at 10:08 AM
Never play Battleship against a kid who can’t count
February 5, 2025 at 10:08 AM
One thing I'll never forgive Google engineers for is their habit of choosing clever, cutesy, horribly opaque names for EVERYTHING
February 5, 2025 at 10:08 AM
Early-career engineers - one superpower that can set you apart is honing a systematic approach to hairy problems. The scientific method! 1/
February 5, 2025 at 10:08 AM
Why, yes, that _is_ Mt. Laundry in my Zoom background with a crying baby on its summit. #nonannynoproblem
February 5, 2025 at 10:08 AM
Been pairing with Copilot. My hot take: I much prefer pairing with humans 🕺
February 5, 2025 at 10:08 AM
Second in a 3-part series: in which we model a financial transaction system with TLA+ to track down a sneaky concurrency bug.

https://www.g9labs.com/2022/03/12/what-s-the-fuss-about-formal-specifications-part-2/

#tlaplus
What's the fuss about formal specifications? (Part 2)
In which we debug a production bug (loosely based on a re...
www.g9labs.com
February 5, 2025 at 10:13 AM
Me speed reading every title on O'Reilly before I lose ACM access
February 5, 2025 at 10:13 AM
Couldn't miss a chance to drop some Boyz II Men lyrics in this tech spec 🎙
February 5, 2025 at 10:13 AM