Hey @claudeai, you made a product that writes code and have billions of $$, yet you can't transfer my account, subscription or credits to a my new email OR give me a refund on my purches credits. EVEN THOUGH I will simply resubscribe and buy them on the new account?!
A fight doesn't happen inside a doctrine. Neither does building your business. Mastery isn't obedience or rebellion. It's merging what you know with who you are and the reality in front of you.
seanshadmand.com/2026/10/14/…
Sean Shadmand retweeted
Designer-made agent skills are everywhere this week.
This is the context and the checklist I'd give my agent before installing one.
The skills going around:
- elayadesign/ai-design-skills on GitHub, by @elayadesigns.
Two skills. landing-page-design asks intake questions, then builds a landing page with its own type, spacing and motion system. design-to-motion turns a static design into an animated showcase, exported as MP4 and GIF. Works in Claude Code, Cursor and Codex.
- Leonxlnx/taste-skill by @LexnLin. Skills that give coding agents taste: art direction, reference images and a browser check.
- Nutlope/hallmark, made by Together AI. An anti-slop design skill that checks the page against AI tells, like an eyebrow above every heading.
Before you trust a design skill:
1. Read the SKILL.md yourself. A skill is just a file.
2. Check the credits. Does it name the skills it borrowed from?
3. Check what it runs. Writing files is low risk. Installing packages or running scripts needs a closer look.
4. Try it on a throwaway branch first.
5. Find the section where its taste lives (fonts, colors, spacing, motion) and swap in yours.
Prompt to paste into your agent:
"Read the SKILL.md at [repo URL]. Tell me what it does, every command it runs, what it credits, and which parts conflict with my design system. Don't install anything yet."
The first skills I installed this year, I rebuilt to my own taste right away.
What's the first thing you check before installing a skill?
Sean Shadmand retweeted
Making your AI agent a user of your design system.
When you ask your coding agent to build a screen, it's supposed to read your components and your docs to decide how and what to build. It needs four things from your design system.
Review time is the new build time. Agents made "can we build it" cheap, so the expensive question is "should this exist at all." That's still a person's call.
Agents ship faster. Judgment still decides what is worth shipping. @bentlegen on staying human in the loop.
Apollo's Torsten Slok: since 2022, Mag 7 revenue per employee went from about $550K to nearly $800K. The rest of the S&P 500? Flat.
The tech isn't the bottleneck. Habits, trust, and someone accountable for the agents are. You can't adopt AI cold turkey.
benzinga.com/markets/tech/26…
A weekend side project: I wanted everything my agents do in one place, just to track effort and usage. So I built Penopta and open sourced it. It's lightweight, has a Mac app, and splits work by plans and projects. When I lose something, or need my time on a plan, it's there.
penopta.com
Code: github.com/sshadmand/penopta…
I think it sort of is a lot like beacon.sh/pricing. All I would have to do is turn the shared history into skills I suppose.
Feeling like you're not going fast enough? One more feature. One more customer type. One more exception. One more "why not?" You can't address every problem at once. Lighter is what makes you fast.
seanshadmand.com/2026/06/19/…
One researcher plus AI agents pretrained a frontier coding model in 3.5 months. Agents wrote 99% of the code; the human decided what to trust. It's been true for decades: when building is cheap, deciding is the work.
mirendil.com/news/pretrainin…
seanshadmand.com/indecision-…
Attacks and fixes now move at machine speed, but deciding to patch, taking a site offline or changing a policy still moves at human speed. Technology has never been the slow part of change. People and institutions are, and days like this show the gap.
Replying to @cognition
@cognition says many coding tasks are “intelligence saturated”: once a model does the job well, cost and speed matter after that. People don’t always need every possible outcome; they need reliability ans speed. Sometimes the best response is no AI at all.
businessinsider.com/cognitio…
seanshadmand.com/2026/10/06/…
There has always been a delta. The delta was never the hard part. The bottleneck has always been the "if you so desire" part, and knowing to not think everything you can think up is needed. Cutting maketh the brand.
My blog posts are now a podcast. Short reads, 3 to 8 minutes each, in my own voice. Listen on Spotify: open.spotify.com/show/58JhEu…
Sean Shadmand retweeted
I build UIs for AI agents: MCP Apps, generative UI, and design systems that agents can actually use.
What I'll be sharing here:
• Agentic UI content, demos and deep dives
• Open-source starter kits
• Lessons from shipping agentic UI in production
At React India (Oct 29–31) I'm giving a talk on building a Design System MCP Server, and running a hands-on workshop on MCP Apps.
Follow along if you're building for the agent era.
Sean Shadmand retweeted
ai slop in your interface is a design system problem
at rerun, every screen pulls from the same design system: the app, the landing, emails, whatsapp messages, og images
everything looks like us, everywhere
we have pages that do nothing but display the whole design system
every component, every state, every view
when i want to test an idea, i spin up a "render lab"
one temporary page, one experiment, thrown away right after
that's where we iterate together
prototype, compare, break stuff, until the ui and the ux actually feel right
and teammates who don't code can do it too
they iterate, they build components, i just implement them
real example: we swapped our flat agent blocks for little figurines
~3 files changed, and it showed up in the app, the landing, the og images and our x visuals
in 2026 code is worth almost nothing
iterating fast inside your design system and your app's ergonomics is how you ship banger interfaces
This is the right direction: explore UI with your agent using the real components and styles already in the repo. The other half is keeping every variation true to the design decisions the team already made, so exploring never turns into drift.
new skill: create-canvas.
last week I made a skill to put every component from a repo on one canvas. I started using it with my agent to create variations, compare ideas, and iterate visually without leaving the codebase.
so I made a new skill around that workflow.
instead of moving your work into another tool, bring the canvas into your codebase and use it to explore UI with your agent, using the real components, styles, and project context already there.














