Patrick Lu (@lightninglu10) header
Patrick Lu (@lightninglu10) profile picture

Building agents that automate everything in your company: https://teapawt.org/t.co/hhaxPjaLDL / ycombinator fellowship / top 80 NA FIFA player

Berkeley, CA
Joined October 2012
Photo by Patrick Lu (@lightninglu10)Photo by Patrick Lu (@lightninglu10)Photo by Patrick Lu (@lightninglu10)Photo by Patrick Lu (@lightninglu10)Photo by Patrick Lu (@lightninglu10)Photo by Patrick Lu (@lightninglu10)Photo by Patrick Lu (@lightninglu10)Photo by Patrick Lu (@lightninglu10)Photo by Patrick Lu (@lightninglu10)Photo by Patrick Lu (@lightninglu10)
#throwbackthursday Back before I built software and my dream was to make it as a pro soccer player
1
1
12
4,791
Yeah don’t use fable to generate work outputs. It’s not meant for that kind of work. It’ll cost you 2.5x and 3x as long vs using Opus. Have Fable sketch out what direction your assets should be in and have Opus generate those assets. This is the way
How are people getting Fable to generate such high quality graphics without using third party assets? I've tried and end up getting the same Roblox graphics no matter what I say or reference images.
1
51
All this talk about cloud sandboxes. Have y'all used sandboxes in the cloud? My entire workload would cost me $5k+ a month. Instead, just own your compute. $5k buys you a 128gb macbook. It pays itself back in 1 month. 5 computers, it's a no brainer. Only use rented clouds if you need to, otherwise own your compute.
44
We just totally changed the way we do design @ CodePress No more design handoffs / wireframes / mockups. Our agent's become our designer :D We're open sourcing the code that lets us do this: github.com/codepressdev/desi… It'll work in codex or claude code but if you want it no setup, check out codepress.dev/app -- our orchestrator app where we use any model we want
3
5
123
We just killed Figma Wireframes, mockups, design review, handoff: gone. We merge 100+ PRs a day per engineer, 24/7. Our agent is now our "designer". It draws every screen from our real components -- every state, every error. We review, approve, and it ships what it drew Today we're open sourcing the design kit that makes it work
1
5
359
The design kit works with Claude Code, Codex, or any agent. We have skills to help you generate a Storybook if you don't have one, and once you do we capture real app pages + components so your agent composes designs from your real app. github.com/codepressdev/desi…
1
49
China will never win the AI wars because they don’t care about the last 10%. The way society is structured, you don’t get rewarded for going the extra mile. I was in China this past May, and this reflected in every aspect of society. A mall that someone spent millions to build, only to skimp out on the final finishing. A sidewalk that stopped right before it connected to another one. A grand building / office only to use cheap materials so it looks terrible after 10 years. Everyone in the US knows the last 10% is everything. It’s the difference between AI actually doing the job vs giving you shitty 1st draft AI slop. That craftsmanship and quality actually matter. That’s why China won’t win the AI wars. They only care about “good enough”
25
The only way to move at agent speed is to give up control to agents. If you’re needing to weigh in on every decision, you’re bottlenecking everything. Let them ship. Let them figure it out.
1
33
Build your own software factory with codepress.dev You can customize everything about it, including your own UI / brand + domain, where your agents run, and how the factory behaves. We give you a base, and also run cloud agents / cloud infra where you don’t want to deal with it
You're really going to want your own software factory
59
AI this, AI that, whenever I’m back in the south all I want is a good ol’ steak and eggs from @WaffleHouse
1
3
75
And yeah ofc I got the waffle
26
You’re getting lapped 15x over and you don’t even know it Everyone knows speed is the #1 pillar of startup / business success. Speed of implementation = speed of experimentation = finding pmf faster This is a PSA that you could be shipping much much faster
Bookmark this: our entire system for how we 14x'ed our output from May to October 👇
1
53
Man, if you don’t have agents teams working over days, you’re really missing out on the productivity / alpha you can have. One project’s been working for me for 3 days, wouldn’t have been able to ship this previously unless I put 100% focus on it 27 PR’s shipped, 2 to go 😙
1
47
Madman is load testing, making sure edge cases are handled properly, and trying to break it / find a vuln in it all at the same time
24
If you’re not shipping this much, then what are you doing?
Bookmark this: our entire system for how we 14x'ed our output from May to October 👇
1
2
113
Bookmark this: our entire system for how we 14x'ed our output from May to October 👇
4
5
1
14
748
That's over 500 projects last month :D
49
I had to buy my own server to support the increased shipping speed we had. We’re shipping 13x faster than in May. 5 months, 13x speedup. The world isn’t ready for all the compute we need for agents
This week we didn’t see an uptick in shipping velocity. We stalled due to CI capacity issues with our provider, and to make things worse, even our cloud providers have run out of capacity to host our runners. VCs are getting excited about funding faster/cheaper runners while the reality is that constantly optimizing and rewiring CI/CD has become the name of the game. Once you solve the agentic validation and code-review bottlenecks, you still have plenty of engineering and janitorial work to keep the factory humming.
1
121
xApple x @a16z Someone here’s gonna build the next trillion $$ company
3
112
All this talk about PR’s when your amp orb is just a PR with a different skin. We’re not using PR’s cuz we have other devs review them. It’s because we don’t need to build a new UI layer for the agents to talk to each other. They post on the PR to record their decisions. They use “ready to review” as a way to know when to do adversarial review. Saying “you don’t need a PR” is stupid because yes you don’t need a human to review it, but you still need all the UX layers around it, even if you mask it as something else.
In a world where every developer is now wielding agents. You can measure each phase in the SDLC. Can guarantee the PR (reviews, approvals etc) is the least efficient step in the process. This is why we don’t use PRs at Amp.
1
1
2
424