Author | Lightnews

Andrew Hao

@andrewhao.com

73 followers 170 following 1.7K posts

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

mastodon.online

Posts Media Videos Starter Packs

Andrew Hao @andrewhao.com · Oct 4

Needs very small checkpoints here and there. Needs tons of context and checking in. Not unlike a human.

Can't be afraid to throw away work and start over if it's not working. You still need, as an engineer, mastery of the craft and a vision for what you're building toward. AI won't solve that!

Andrew Hao @andrewhao.com · Oct 4

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.

Andrew Hao @andrewhao.com · Jul 10

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.

Andrew Hao @andrewhao.com · Feb 9

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!

1 3

Andrew Hao @andrewhao.com · Jan 9

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

Andrew Hao @andrewhao.com · Jan 8

2025's goals are to create more:

🎵 Music
✍🏼 Blogging
📀 Side projects

Andrew Hao @andrewhao.com · Jan 8

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!

1 1

Andrew Hao @andrewhao.com · Jan 2

I'd listen! I have no idea how other parents make all the other parts of their lives fit.

Reposted by Andrew Hao

Kent Beck @kentbeck.com · Jan 2

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

4 18

Andrew Hao @andrewhao.com · Dec 4

Thanks for doing the good work Rich!

1 1

Reposted by Andrew Hao

Yoel Roth @yoyoel.com · Feb 21

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.

Casey Newton @caseynewton.bsky.social · Feb 21

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...

Even more interesting, though, given the current interest in novel risks posed by AI, is that a last-generation OpenAI model performed almost as well as the foreign agents. Here are study authors Josh A. Goldstein, Jason Chao, Shelby Grossman, Alex Stamos, and Michael Tomz:

GPT-3-generated propaganda was also highly persuasive, and 43.5% of respondents who read a GPT-3-generated article agreed or strongly agreed with the thesis statement, compared to 24.4% in the control (a 19.1 percentage point increase). This suggests that propagandists could use GPT-3 to generate persuasive articles with minimal human effort, by using existing articles on unrelated topics to guide GPT-3 about the style and length of new articles.

7 100 240

Reposted by Andrew Hao

Jay 🦋 @jay.bsky.team · May 1

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.

Mycelial Ghoul @backwoodsfairy.bsky.social · Apr 29

Tell us a true story from your life that sounds made up.

66 290 1.5K

Andrew Hao @andrewhao.com · Nov 19

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 😭

Andrew Hao @andrewhao.com · Nov 19

I love data scientists. They are amazing. That is all.

Reposted by Andrew Hao

Julia Evans @b0rk.jvns.ca · Nov 23

git discussion bingo card

a bingo card with

* WTF is detached HEAD state
* I hate git
* git's design is so elegant
* I've used git for 10 years and I have no idea how it works
* commits are immutable snapshots
* just use magit
* git is a directed acyclic graph
* you have to understand the linux kernel dev workflow
* mercurial is better
* git is not github
* subversion was so much worse
* I only know 5 commands
* a branch is just a pointer to a commit
* the CLI is badly designed
* you should just read Pro Git
* rewriting history is bad
* just spend 15 minutes learning git's internals
* something about "porcelain"
* merge sucks, only use rebase
* rebase sucks, only use merge
* I just do not care how git works
* content addressed storage
* subversion was better
* something about Linus Torvalds
* I just delete my git repo if I mess it up

21 87 250

Andrew Hao @andrewhao.com · Nov 19

Congrats you two!!

Andrew Hao @andrewhao.com · Nov 19

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!