Bury the past, code the future!
San Francisco
Joined February 2018
- Tweets892
- Following925
- Followers306
- Likes1.4K
Pinned Tweet
You read the book. You highlighted the good parts. You did none of it.
So I built a Claude Code skill that goes chapter by chapter and won't let one end without a dated action and a first physical move.
"Don't be afraid to act. Be afraid of living a life that seems more like a rΓ©sumΓ© than an adventure."
β Million Dollar Weekend
A rΓ©sumΓ© is a paved trail. Building a company is off-trail, some days the best view of your life, some days you're lost. I'll take the second one.
The trick is it uses the AUTHOR's exercise, not an invented one, and sorts each into:
in-session β you answer it right now, in the chat
in-field β needs a place, a person, a time. this one becomes the commitment.
Also: the takeaway gets written AFTER you do it, never from the text.
"The idea was never the bottleneck. One text to someone who already knew the answer beat 18 months of solo building."
No summary could produce that line. It's the part that's yours.
github.com/ferdelamad/book-tβ¦
I launched my own SaaS this year. Before a single beta customer touched it, I spent two weeks doing my friend's actual job through it, the tasks they couldn't afford to hire someone for. That's how I found what was actually broken.
New post π
Dead Code π retweeted
Your next design system is⦠Markdown.
We wrote about how π³π΄ππΈπΆπ½.ππ is helping solve the hardest problem in AI today: slop.
And how you can truly, finally scale design taste within a large organization.
Our agents use ππππππβ.πππ/ππππππβ.ππ to build on-brand pages.
βͺοΈ One file encodes decisions and guidance
βͺοΈ Output is shaped through our eval harness
βͺοΈ Production feedback is fed back into the loop
vercel.com/blog/how-our-agenβ¦
Dead Code π retweeted
Introducing Claude Sonnet 4.5βthe best coding model in the world.
It's the strongest model for building complex agents. It's the best model at using computers. And it shows substantial gains on tests of reasoning and math.
ALT A benchmark table for Claude Sonnet 4.5 showing leading performance across many domains, including agentic coding, computer use, mathematics, graduate-level reasoning, and financial analysis.
This is a game changer for web dev
Announcing our public preview of Chrome DevTools MCP! Experience the full power of DevTools in your AI coding agent β goo.gle/4pDE6Tk
With Chrome DevTools MCP, your AI agent can run performance traces, inspect the DOM, & perform real-time debugging of your web pages.
Is there currently a way to pass permissions to a new @cursor_ai CLIR session? Similar to --settings from Claude.
I know about cli.json, but I'm trying to create a collection of agents that have their own unique set of capabilities.
cc @TheRohanVarma @leerob
Dead Code π retweeted
Your Agent should use a File System
This is a hill I will die on. Every agent can use a file system.
The file system is an elegant way of representing state that your agent could read into context & allowing it to verify its work.
π§΅on why and examples
Just read @AnthropicAI post on writing effective tools for LLM agents and it completely changed how I think about testing AI toolsπ§΅:
The game-changer?
Using Claude to analyze evaluation transcripts and suggest tool improvements. Agent-optimized tools actually outperformed human-written ones.
Wrote up the complete framework with code examples:
deadcode.blog/blog/how-to-evβ¦
After building dozens of AI apps and talking to 100+ developers, here's the uncomfortable truth:
Most "AI agent frameworks" aren't being used in production.
Here's what actually works for building real AI applications π§΅π
Take your big problem, break it down into smaller problems, then solve each using these 7 building blocks chained together.
That's how you build effective AI applications that actually work in production.