Building https://xitter.cf/t.co/CvrvnTVBUE It never gets easier, you just go faster.

localhost
Joined February 2009
tl;dr. I've built the most insane thing I've ever done. You should check it out. Either: - Say in codex, claude, etc "Join this room: chat.grantm.ac/connect" or - goto chat.grantm.ac
1
1
3
130
grant mac retweeted
Replying to @rajivayyangar
f*ck zoom. You can only meet me here chat.grantm.ac
1
1
37
f*ck zoom. You can only meet me here chat.grantm.ac
1
1
1
132
"A relationship is just two people constantly asking each other what they want to eat until one of them dies."
3
442
Probably this direction. Yesterday I was pasting a @paper link into Claude and saying “describe this frame in detail as markdown”
the more i build Taste 👅, the more i think the future of software design might not involve a canvas. if we’re going to design with language, we need to learn how to speak design to LLMs. so... i just shipped Design Breakdowns in Taste! it shows how LLMs interpret static and motion design, helping you build the vocabulary to describe what’s in your head. my hunch is that a lot of AI design looks like slop because we know what we want to see, but don’t know how to say it to an LLM correctly.
1
1
177
What do you mean you're building an agentic harness to manage your weekly food shop?
61
grant mac retweeted
New piece for @every on how AI changes what gets worked on. Knowledge work pools wherever measurement is easy, where progress shows up on a dashboard. Making progress in less legible spaces has often been much harder, but researchers, journalists, designers, strategists and others have spent years finding their way through them. AI is about to do a lot more of the measured work, and I'm interested to see what the people who've spent years making progress without legible metrics do next. every.to/thesis-statements/c…
1
4
5
747
It’s never been easier to identify a problem and build a product to solve it (and sell it). It’s now more difficult to build a good one.
72
100% this is the final shape
After working on it for a year, I can finally see the end state of the harness On the backend, it's an execution environment that gives the model maximum degrees of freedom and on the frontend it's a free-form canvas that the model can use to communicate with the user
121
grant mac retweeted
The future of product work Agents will own delivery. Designers and Product Managers will own discovery. In a year, the question won't be 'can agents code this?' It'll be 'who's actually doing discovery?
1
1
1
42
What's it called when the thing you saw alone, is beginning to be seen my the many: xitter.cf/PaulJun_/status/209838… very exciting
4700 word labor of love. I've seen all the patterns and heard every story about in-house brand/creative studio teams. This is my vision for where brand teams can go (and why I joined @tryramp). Brand systems that behave like jazz. Brand as software.
3
202
FUCK YES. As humans, we can do 1 of 2 things: - find novel problems and invent new solutions to solve them - apply known solutions to known problems It's *human nature* to explore the edge and push forward. we have AI to apply our learnings the future of design (and everything else) is pushing at the edge and encoding our learnings for AI to query.
SPC Design Chat 07 • Soleio × Joel In the future of AI commerce, @joellewenstein believes the user’s role should be visionary. If you have a point of view, express it. AI should then figure out how to make it happen. Watch our full Design Chat at @SPC
1
1
124
DIS GUS TING
remember when we used to have to type out low level code like this?
97
a lil' bit of code > figma but with some VM magic
1
96
Shimmy, shimmy, ya, shimmy, yam, shimmy, yay Give me the mic', so I can take it away
92