andyrewlee (@andyrewlee) header
andyrewlee (@andyrewlee) profile picture

agent orchestrator connoisseur

San Francisco, CA
Joined April 2009
Photo by andyrewlee (@andyrewlee)Photo by andyrewlee (@andyrewlee)Photo by andyrewlee (@andyrewlee)Photo by andyrewlee (@andyrewlee)Photo by andyrewlee (@andyrewlee)Photo by andyrewlee (@andyrewlee)Photo by andyrewlee (@andyrewlee)Photo by andyrewlee (@andyrewlee)Photo by andyrewlee (@andyrewlee)Photo by andyrewlee (@andyrewlee)
Your numbers are wildly incorrect
1
55
Presenting my newest AI short film: DON'T BLAME THE COMPUTER. Made entirely with AI - about why that's not the point.
4
1
7
743
andyrewlee retweeted
Replying to @ryolu_
this actually makes me quite sad to read, I'm excited for what interfaces you will build and agree that terminals are not the most mass market coding interface but they are the tool of choice for the best engineers in the world for a reason and it's not romanticism
6
1
72
6,069
3rd Place - Checkfu 🥉 Checkfu creates printable black-and-white Kindergarten worksheets and coloring pages. Each page is represented as an 8.5×11 node containing a single image, which can be refined with a prompt to generate a child variant. Pages can be exported as PDFs for saving or printing. @andyrewlee
3
4
1
6
754
3/ Checkfu Create printable black-and-white Kindergarten worksheets and coloring pages. Each page can be refined with a prompt to generate a child variant. Pages can be exported as PDFs for saving or printing. @andyrewlee
3
1
26
4,868
andyrewlee retweeted
Vibe coding a platform built on irreversible transactions will surely not cause any issues
~40% of daily code written at Coinbase is AI-generated. I want to get it to >50% by October. Obviously it needs to be reviewed and understood, and not all areas of the business can use AI-generated code. But we should be using it responsibly as much as we possibly can.
68
41
5
2,291
126,022
Makes sense. We've done this for a year now. It changes the challenges you do, shifting away from rote recital of idioms to more systems thinking, rubrics about "why did you implement X instead of Y" rather than "did you do X right".
4
1
1
24
3,993
Run from it all you want This is the new reality
Just saw a great PR come in. Killer feature, beautiful description, commit messages, comprehensively tested code (40% of code is tests). Ask the engineer what % of this is AI. The answer: 100%
3
22
3,533
andyrewlee retweeted
I’m so pumped that y’all are discovering the magic of Elixir
code like this makes me realize that elixir is just so objectively superior i can match messages, in the function definition, and just match the prefix... this is just such an annoying thing to do normally
7
6
206
28,494
Feature parity is just another way to say “we don’t need to exist”.
21
19
7
265
35,928
2020: “LLMs can’t do math.” 2025: LLM wins gold at the IMO
63
175
31
4,178
306,858
andyrewlee retweeted
QA is the new SWE
50
34
13
627
106,339
phoenix.new is publicly available! I can't wait to see what folks build and where this goes! - remote AI runtime for full-stack app building, github workflows, etc - Full in browser IDE and xterm, with root - Preview URLs for any port fly.io/blog/phoenix-new-the-…
27
98
19
480
69,214
andyrewlee retweeted
I really like the term “context engineering” over prompt engineering. It describes the core skill better: the art of providing all the context for the task to be plausibly solvable by the LLM.
346
861
298
8,568
2,091,164
andyrewlee retweeted
ask your team: "what happens to our product when models get 10x better?" if the answer isn't "we become 10x more valuable" you built for today's failures instead of tomorrow's capabilities time to pivot or die
13
14
1
207
14,842
Today, we're announcing plans to make VS Code an open source AI editor. We believe AI development should stay true to VS Code's core principles: open, collaborative, and community-driven. Let's build the future of software development together. aka.ms/open-source-ai-editor
365
2,080
867
14,115
1,869,840
andyrewlee retweeted
the best measure of a codebase's quality will soon be how productive an LLM can be inside of it without human intervention
12
14
9
219
15,702
A famous person wore a secret hidden wire while speaking with me about AI, and I want to talk about it. Backstage at a recent AI event, I was talking with someone about private views on AI investments and strategy. We were both unfiltered. At one point, I joked, “We should’ve recorded this - we’re dropping gems!” He tapped his chest. “Oh,” he said. “I’m always mic’d up.” I laughed. “Wait, are you wearing a wire?” “Yes.” He had a secret mic under his shirt the whole time. No disclosure. No consent. Just recording. I felt completely violated. Not because I said anything wrong, but because I didn’t choose to share it. As I think about the future - AI-powered glasses like meta rayban, smart earbuds like airpods, recording necklaces like limitless - these are tools that serve their owner, while extracting data from everyone else nearby. The benefits are asymmetric. The consent is invisible. We talk a lot about surveillance capitalism and corporate tracking. But I do think we’re also going to see both intentional and unintentional “interpersonal surveillance” (honestly it’s already happening now with friend search, purchases, and Bluetooth/WiFi tracking, and with people constantly filming everything in their lives on their phones). The ones wielding the tech get a boost. Everyone else is turned into passive data sources. If always-on wearables become the norm, I’m thinking we also see a new kind of social friction. People will retreat. Trust between friends will erode. And ironically, those recording everything might find themselves with fewer conversations in real life. The cost of ambient AI might be wider than we first thought.
210
227
108
1,841
325,906
andyrewlee retweeted
It seems to me that AGI would mean the end of prompt engineering. Moderately intelligent humans can figure out what you want without elaborate prompts. So by definition so would AGI.
584
273
155
6,108
513,038