Sabitlenmiş Tweet
Tom Krcha
8.9K posts

Tom Krcha
@tomkrcha
Chief Vibe Officer at https://t.co/YJUz1TqYvM ✏️
San Francisco Katılım Ekim 2008
1.6K Takip Edilen20.3K Takipçiler
Tom Krcha retweetledi

Did we get a chance to tell you about Mesh Gradients in Pencil.dev yet. Yeah, they are really cool. Try them out. Go to fill -> mesh gradient. Or ask agent to generate one for you to get started.

English

@tomkrcha loving Pencil. can see this being the future of designing with AI agents. thanks for building it.
seeing this error today. and didn't find a way to debug. asked claude but it gave up. what is code 1 and how to solve?

English

Hi Tom, incredible tool. I was showing swarm mode to my wife last week. She runs social media & AI tools at her gaming studio. This morning I found her giggling at her desk, she grabbed her laptop, shoved it in my face, and showed me three websites she was building. (That's never happened before.)
You're building something genuinely special. I work at the intersection of growth and creative. Would love to find a way to help.
English

Big update: I’m starting a new company.
6 months ago, @_bschmidtchen and I made a bet. What if entire worlds could be generated on the fly, pixel by pixel?
World models are the next platform shift, and we saw it coming.
Since then, we’ve:
- secured major contracts across media and physical AI industries
- assembled a team of 10+ from Apple, Meta, Google, Adobe & Microsoft
- raised from top-tier investors
More details soon. We’re scaling fast, and hiring now.
Come build with us: reactor.inc
English

@benjitaylor 100%. Instead of convincing people to come to you, just join them on the way. One of our core principles building pencil.dev and why we started in Cursor as an extension + inside a git repo, later with deep Claude Code/Codex integration, etc.
English

Taya is jewelry first. Surveillance never.
We just raised $5M led by @MaCVentureCap, @fcubedvc, a16z @speedrun.
Directional mics and voice isolation capture just you. Single-player mode. Privacy isn't a feature, it's an architectural decision.
TechCrunch exclusive in the comments! 👇

English
Tom Krcha retweetledi

Tom Krcha retweetledi

I've been using @pencildev for all of my design work and honestly, it's wild... it's like Figma and Cursor had a baby.
You type what you want, AI agents build the design and you get to watch them work in real time.
What I love:
- Component-based designing
- Agent swarms that design in parallel
- Full Figma-style controls when you want them
- Free to download — just use your existing OpenAI/Anthropic subscription
I use it for all my AI design work now. Not affiliated, just genuinely impressed.
I'm getting the founder @tomkrcha on for a podcast recording soon. What questions should I ask him?
English

@michalblaha 🙏🙏🙏 working around the clock to make it as smooth as possible. I’ll ping you.
English

Just tried Pencil @tomkrcha and it's really good. Thx!
Claude integration needs a little but polishing but no big deal
English
Tom Krcha retweetledi

Spent an hour with pencil.dev and I'm sold. If you're an engineer who has strong design opinions but zero design skills (ie. me), this is your tool.
Free, runs on all platforms, uses your own claude sub with no setup, and the built-in prompts actually work great.
English

Watching 6 AI agents design an app together genuinely blew my mind.
In my next episode, @tomkrcha (CEO of Pencil) showed me:
→ Swarm mode: Spinning up 6 design agents
→ How it all works under the hood (JSON?!)
→ How to design inside Cursor & Claude Code
📌 Subscribe to get the full episode tmr: @peteryangyt?sub_confirmation=1" target="_blank" rel="nofollow noopener">youtube.com/@peteryangyt?s…
English

The hard part about LLM failures is that their outputs rarely look like failures.
The demo “works.”
The output sounds coherent.
The user actively uses the product.
And your dashboard looks normal.
Meanwhile, the system can be wrong, unsafe, or quietly driving up token spend. And you won’t notice until the damage adds up.
Prompts often serve as business logic (policies, safety, and product context). But many teams ship them without the basics, such as versioning, reviewable changes, end-to-end traces, and eval gates.
In production, it doesn’t crash. It degrades via wrong answers, policy misses, and surprise spending.
No crash. No error. No alert.
I cover this exact issue in my @Stanford CS 224G guest lecture on AI Observability and Evaluations.
Here are the core ideas:
• If you only log the final output, you’re guessing. Full traces show where it broke.
• Evals are feedback loops. Use clear pass/fail criteria tied to outcomes.
• Run evals continuously on production traces and don’t wait for support tickets.
The moat isn’t prompt cleverness. It’s a measured improvement.
Full lecture + blog below 👇
English

fun lil' weekend project
macOS app called shiftcut
load up a library of sound effects and it automatically detects the sound spikes and creates keyframe animations based on the audio signature
adjusting duration affects the sound and the transition speed together
generate transition effects with claude code
all UI functions are also CLI commands
export as black and white .mov with sound baked in
use with luma key in any video editor (see FCP at end of video)
English

"These cursors seem like a small touch, but it's the first time I've seen AI feel human."
Here's my new episode with @tomkrcha (CEO of Pencil) where we covered the most mind blowing AI design tool I've ever seen.
Tom showed me:
✅ Swarm mode with 6 AI agents designing an app at the same time
✅ How to load a full design canvas inside Cursor and Claude Code
✅ From design to working website in one prompt
Some quotes from Tom:
"We launched 2 weeks ago and now have 100,000+ users. Craft and care still matter."
"Most vibe coding platforms are too linear. You should be able to explore 20 variations with agents."
"All this magic on screen...behind the scenes, the agents are just writing JSON."
📌 Watch now: youtu.be/w4RY7PnfRU8
Thanks to our sponsors:
@linear: The AI agent platform for modern teams linear.app/behind-the-cra…
@Replit: From 0 to full stack app in 2 min. replit.com/?utm_source=cr…

YouTube
English


@jessperate yep, you need to click on ⚡️ in the chat on the bottom, Mac/Windows/Linux apps with Claude/Codex
English

@tomkrcha Is Swarm mode live in existing product? Not seeing it in the latest update.
English

@petergyang @tomkrcha 6 designers and nobody argued about the font
English

@twlvone @petergyang That, but also, truly parallelizing coding is really hard, parallelizing design, much more realistic, that’s why I am so bullish on vibe designing. From an idea to preview of a full app in minutes. Think of it as visual planning mode.
English

@petergyang @tomkrcha the wild part isn't 6 agents designing. it's that agent coordination actually works now. a year ago multi-agent was a guaranteed mess
English









