Head of Design at Samsara. Tinkerer, gear head, and wannabe YouTuber.
San Francisco
Joined March 2013
- Tweets1.4K
- Following283
- Followers250
- Likes1.7K
"keep iterating until you get it right" is the best way to end a prompt. Claude will cook for hours
1
44
I’ve met about 20 design engineers in the last month and their profiles couldn’t have been any more different.
2
46
Taron Ghazaryan retweeted
Our design system used to be two things: a Figma library for designers + components in Storybook for engineers, with a lossy translation between them.
It’s now one source of truth that designers, engineers and AI agents all read from. Our head of product design @TaronGh on what changed →
We'll probably end up with 3 categories of apps
1. Utilities - camera, maps, music, etc.
2. Entertainment/browsing - Netflix, YouTube, TikTok
3. AI assistants - everything else.
1
57
Every time I see a tweet like this, the output is usually some useless piece of software no one is going to use. Every single time.
Two take-aways from this:
1. Gap is getting wider and wider by the day between people who have learned how to orchestrate these tools for exponential benefit - versus people who use AI like a search engine or have no idea how to use AI at all.
2. This incredibly impressive website was built while yelling at a phone while on a bike: felixrieseberg.com/
I think the fact that 2 is true just further reinforces the raw truthfulness of 1
60
Check out our design team site and how we work :)
At Samsara, our designers shape how physical AI works for the people who keep the world running. They spend time in the cab, the yard, and the shop to get it right. Check out our new design site: samsara.design/
1
38
Ideal builder software.. @figma + @cursor_ai in one package
- Connect any model
- Design system lives in code, not replicated Figma components
- Principles and heuristics .md managed through the tool, tested against code
- The code is the product source of truth
4
3
95
Then someone just rips your idea because the limit in copying software was also engineering.
The limit on software used to be engineering and now it's mostly imagination.
I'd spend a lot more time on the version of my idea that makes people say "wait, what is that?"
Most software looks the same, so anything that looks different gets noticed, shared and remembered.
1
1
60
Not trying to dunk but what’s the point of framer or webflow anymore? Can’t you build whatever you want in Claude/codex?
Watch this at 2x (3.5 min) to learn how to create a design system and skills for your @framer site so you can then easily build new pages that look way better and adhere to your style guide in a way that out-of-the box ai doesn't come close to...
2
1
102
After wasting a bunch of money, I can confirm neither Astra nor Opus 5.5 are good at 3d modeling
1
1
54
Watching the usage limit deplete while building something big feels like driving while the range says 0 mi
1
64
100000%
The job market is a tale of two cities right now. The difference I am noticing between the folks who are inundated with eye-watering seven-figure offers and those who have trouble finding a role is simply the ability to keep learning, adapting, and getting better.
It's a brave new world out there. Gone are the days of sticky notes and following the "right process."
Things to start doing :
1. Get comfortable with code. Like, now. If you missed the boat on trad code, plenty of IDEs to play with. Did you start yet? No?? I'm begging you, please. For me.
2. Build personal projects. What are you building in your free time? What is the real you, free from the shitty design systems that bind you, capable of? This also clearly demonstrates a builder's mindset and agency.
3. Work on your visual design and creativity. I think years ago, "complexity in the work" and "product thinking" allowed us to get away with weaker craft, but no more. Get weird, people. Have fun with your site and portfolio. Make something memorable. A great start would be @joshpuckett 's Interface Craft.
4. Get curious. It's a fucking designer's playground right now, there are so many fun tools out there to experiment with!
100
Unfortunately the first real innovation with AI happened through chat so now everyone thinks chat is the future.
Chat is awful at precision, impractical for consumption, and inefficient for interaction.
It’s great for open ended intent, but that’s a sliver of use cases.
3
172
