Unc. Emo Millenial. Built https://xitter.cf/t.co/CohjpHOKc4. Now AI @cloudflare. 🌊 https://xitter.cf/t.co/ozlNRa7sHR 🤖 https://xitter.cf/t.co/h062f7xmGY 💬 https://xitter.cf/t.co/GfRK9ximUe 🧊 https://xitter.cf/t.co/YzYXNJQQL5
NC, USA
Joined February 2009
- Tweets4.7K
- Following2K
- Followers915
- Likes3.2K
Pinned Tweet
I got tired of XCUITests taking over my screen every time an agent tested my Mac apps, so I built offscreen.
It runs UI tests, the iOS Simulator and GUI apps in throwaway macOS VMs.
Agents queue jobs and get back logs, screenshots and results.
github.com/markphelps/offscr…
Mark Phelps retweeted
Nice to see our transcription model, Voz, in Ramble, a new Mac app from @itsmarkphelps. Point at something on your screen, talk it through, and your coding agent gets screenshots, the UI details and a timestamped transcript in one file.
Voz runs on the Mac, so your audio never leaves it.
ramblemac.com
I was about to launch Ramble as a paid app. Then this went viral last night, and it's basically the same.
Hard to compete with free, so Ramble is free now too.
Point at something on screen and talk. Your coding agent gets screenshots, a transcript, and the UI details from macOS accessibility.
ramblemac.com
Two tools I shipped this month started the same way.
Slow, expensive macOS CI runners. Now a skill that sets up self-hosted ones.
Agent UI tests taking over my screen. Now offscreen.
I didn't plan either one. I got annoyed, then built the fix.
🏃`npx skills add markphelps/skills --skill github-self-hosted-runner`
🖥️ github.com/markphelps/offscr…
idk how he does it. Jordan keeps shipping banger after banger like every day
write-judge is a small Pi extension that sits in front of every edit / write.
Before the change lands it asks one narrow question of @Cloudflare’s Clef decision model:
Does this change delete or weaken a test, assertion, check or guard?
github.com/acoyfellow/write-…
Mark Phelps retweeted
write-judge is a small Pi extension that sits in front of every edit / write.
Before the change lands it asks one narrow question of @Cloudflare’s Clef decision model:
Does this change delete or weaken a test, assertion, check or guard?
github.com/acoyfellow/write-…
If you manage engineers: what's the recurring chore that eats your week and isn't building or coaching?
Genuinely curious. Not selling anything.
Been thinking and experimenting with this pattern a lot lately.
AI agents that can modify your site / app while it runs.
Anyone can have a custom version of any app.
Jordan cooked with this. One prompt and 2 minutes later here’s mine
skin:
an app that rewrites itself while you use it, and can't brick itself
a @pidotdev Pi Durable agent, which on every change:
bundled in a Durable Object,
booted in a Worker Loader isolate,
checked,
then live fails the gate
crashes in your browser? auto rollback
@EffectTS_ v4 + @foldkit + @tailwindcss, on @Cloudflare Workers
Sign in with Cloudflare: every model call runs on your account (not mine, sorry <3)
skin.coey.dev
Its also not limited to GUIs or tests. Can run anything on the VM or on your host.
In host mode it waits until you are afk and you can take back your computer at any time
Install: `brew install --cask markphelps/tap/offscreen`
Built on Tart. Open Source, so issues welcome.
I swear since switching to using git worktrees my local disk fills up like 5x as quickly.
I need a better solution chat
“So eventually the products develops a character through what its makers pay attention to, what they decide not to do and what they keep working on after it technically works.”
— Someone behind the software
karrisaarinen.com/posts/some…
via covereader.com
