
AI Product Architect, @materialbank a.k.a. @siteinspire. Work → https://xitter.cf/t.co/2ZZiVWOU3l Instagram → https://xitter.cf/t.co/OBlDFX4dS6
London
Joined December 2007
- Tweets6.6K
- Following692
- Followers9.4K
- Likes104
Pinned Tweet
After about 20 years of being indie, I've taken on a full time role at @materialbank and @SANDOW, figuring out what the future of AI-native architecture and interiors products look like.
It's such an insanely exciting time to be in tech, especially in an sector I love, and it's been so good working with @jashsak, @MrGarriock, and Adam Sandow in figuring out what's to come.
linkedin.com/posts/danielhow…
I guarantee however @paper delivers components is going to be superb. It’s my second most used tool after @t3dotcodes right now.
I wanted to share more about Components in @paper to help everyone plan for the upcoming launch. We've been hard at work and we have something truly next level to ship soon.
1. Components dev is going very well. We don't set hard deadlines, we're a "release it when it's good" kind of product, but I'd estimate ~1 month or so. Certainly before EOY.
2. Creating one shared world from your codebase to the canvas and back is the priority.
We don't want to just repeat the abstractions from existing design tools; we want to fix the fragmented workflow between developers and designers. But we also don't want to add cognitive load or engineering concept while you're designing.
Principle: designing with components should feel familiar, designer-first, and then also happen to align to your codebase effortlessly. No tool has ever pulled this off yet.
3. The initial components release will be mature and include: slots, variants with props, React in, React out, and full agent support. Components from your codebase can also run real code, so components can have logic, dates, and math.
4. Agents can read and write components quickly, accurately, and with low token usage. To your agent, components in Paper just look like React, which means they already know exactly how to use them.
I'm so excited to share Paper components with the Paper community. Thanks for your patience as we do the R&D necessary to ship something next level 🙏
Daniel Howells retweeted
Introducing tiny.london
A massive isometric pixel map of London, based on amazing NYC demo by @_coenen from 9 months ago
My agent created the dataset, finetuned qwen image LoRA adapter, did a multiday inference run and built an efficient tile viewer!
I wanted to share something I built over the last few weeks: isometric.nyc is a massive isometric pixel art map of NYC, built with nano banana and coding agents.
I didn't write a single line of code.
Daniel Howells retweeted
Just realized people haven’t been using sparkle icon for AI anymore. That’s progress if we’re assuming all things have AI behind the curtain. But kudos to whoever came up with that icon in the first place - bc boy, it was a fever
Truly brilliant
Introducing tiny.london
A massive isometric pixel map of London, based on amazing NYC demo by @_coenen from 9 months ago
My agent created the dataset, finetuned qwen image LoRA adapter, did a multiday inference run and built an efficient tile viewer!
A good interview question: do you think the Tesla roadster looks good
If they get the answer correct they get the job and a mission s as a company car
I was slow to adopt T3 but Orchestrator in the nightlies has been so, so good. Seamless with two machines and an iPhone, and switching between Calude and Codex accounts seamlessly mid-session, and cross thread comms has been perfect.
Orchestrator V2 is now out on the stable channel in @t3dotcodes! We also thought it was time we drop the Alpha branding, so we call this release 0.1.0 Beta.
Mobile store builds are also updated. Versions 2.0.0 are available to everyone on both App Store and the Play Store!
💯 So true. Seeing ideas shared is like listening to someone describe their dreams now, but seeing something focussed and interesting delivered perfectly e2e is so impressive.
Seeing Fielder’s arc from Nathan for You to being one of the most important and creative powers in Hollywood has been a thrill
Jev is truly astonishing, I’m able to whip out ~30s calls with ~300ms evaluations. BUT it’s not 100% accurate (approx 70% in my tests which is still amazing), and I feel a lot of the demos are really over-reaching/fake.
By November it’ll be perfect and we’ll have to rethink everything.
Al and I agree for once. it's astonishing how awful Astra is with UI. (Still spectacular at 3D though.)



















