cal morris retweeted
Compared to External Agents, Framer MCP is about twice as fast and follows instructions better, because it runs on the same tools as our in-app agent. Setup is simpler too: one approval with your Framer account instead of an API key plus a sign-off for every project, and no CLI or skills to install and keep updated. It also isn't limited to coding agents in the terminal. It works in ChatGPT, Claude, Cursor and any other client that supports MCP, on desktop, web or mobile.
cal morris retweeted
Tip for faster Oxlint on CI: Turn off "warn" rules.
No one reads warnings on CI, so this is a no-brainer. ESLint v10 does this by default, oxlint currently doesn't.
Effect on Framer's monorepo: 100s → 66s, 51% faster.
cal morris retweeted
Get 50,000 agent credits for every client site you plan to switch to Framer. The new Switch program helps you move clients faster and easier, without spending on credits.
Switched? 50% of what we make on your client’s first-year subscription is yours too. framer.link/switch
cal morris retweeted
Skills just shipped @framer so I've turned one of my favourite engineering principles into a reusable skill:
/𝚘𝚗𝚎-𝚝𝚑𝚒𝚗𝚐
It reviews your site, finds a useful improvement and makes the change.
One focused improvement at a time, ready for you to review.
now we need the world first NaN Inf via decision model to make jimbo proud
As promised, here's @Cloudflare's new model "Clef" using NVFP4 quant playing @BalatroGame and winning at ante 8, fully offline on a single RTX 5090!
The last play is mindblowing, very nice combo!
I'm very impressed that it managed to beat the game. Balatro is not a simple game and this model doesn't have a CoT to plan things better like LLMs do. It's not a god-like model, it still makes some decisions that aren't very good but it managed to beat the game in the end, so ¯\_(ツ)_/¯
The response time is not fast, almost 4s on average. The larger the context size the slower it gets. Total playthrough duration was 19m 40s (~3min faster than GLM 5.3 Flash NVFP4 in the sparks).
I had to get close to the 16K token limit trying to squeeze as much info as I could while also attaching a lower resolution screenshot of the game (640x426) thanks to the vision encoder in the model 🙌
This one was especially tricky, as I had to make the options clear and try not to bias and make it an agreeable machine by adding options that I would consider "bad". I can't say that it decided well if there were no bad options to choose from.
The economy seems very different from the GLM 5.3 Flash playthrough, I might give GLM another run with the latest state of the harness to see if it will also spend more money instead of just keeping it.
Also it used a lot of discards, which was quite interesting.
I still think there are a few issues with the harness and I will keep improving it as I try more models.
@CloudflareDev
cal morris retweeted
The Expert program is open for applications again, and it’s now easier to join. All you need is 2 high-quality sites on your profile. framer.com/expert/apply/
✓ 50% revenue share from project transfers
✓ Agent consumes 50% less credits (Pro)
✓ Free edit seat in client projects (Pro)
Been using @theo's T3 Code recently and really enjoying it!
Are there any plans to allow me to chat with Le Faton-Clank (my ChatGPT Dot) inside of T3 Code? 👀
cal morris retweeted
Spent the last week making the @framer Marketplace fast. Like, actually fast.
Core web vitals, real users, p75, start of september → now:
layout shift (CLS): 0.10/0.25 → 0 (desktop was at 0.2 💀)
mobile LCP: 2.8s → 2.1s
Desktop LCP: 1.6s → 1.1s
Stats for /marketplace/templates:
4.9 MB → 2.7 MB
490 requests → 263
63% less data every time you click a template.
The page jumped around while loading. The grid had an empty div above and below it, each with a height, to fake the full scroll length. Every time those heights changed, Google counted it as a layout shift. Now it's padding. Looks exactly the same, but Google is happy. CLS went to literally zero.
Almost a third of all requests were video posters. Most of them were for pages you never even opened. Now it's 5.
We preloaded 7 font files. The page used 2.
SEO got a lot better too. Some pages were serving crawlers a loading skeleton instead of the template data. Now every page hands them the full thing: every tile, the heading, structured data, zero skeletons! No JS needed!
more soon 🫡
cal morris retweeted
I turned our Marketplace guidelines into three Agent Skills for templates, components, and listings. They fix what they can and tell you what they couldn’t check so you can do it.
Behind the scenes on how I built them, plus remix links to try them 🚀
Skills just shipped @framer so I've turned one of my favourite engineering principles into a reusable skill:
/𝚘𝚗𝚎-𝚝𝚑𝚒𝚗𝚐
It reviews your site, finds a useful improvement and makes the change.
One focused improvement at a time, ready for you to review.
As shown in my demo, I've asked the agent to improve something and it made to the historic @framer documentation template I chosen to ensure that each tab includes the Article name. Essential if you're like me and have 20 tabs open at once! 🤯
Grab a copy of the /𝚘𝚗𝚎-𝚝𝚑𝚒𝚗𝚐 skill from my website and let's all share one thing we're doing to our projects this week! 🚀
👉 callum.io/skills/one-thing