
Infra @meta. Exploring https://xitter.gq/t.co/1JBXEnSrpA. Built @medicalscribe_
San Francisco Bay Area
Joined May 2016
- Tweets1.7K
- Following1.4K
- Followers592
- Likes43.8K
Ripped out Node.js from @medicalscribe_ and have been sleeping peacefully since. SPA from now on
🚨 Node.js warns of HIGH-severity security vulnerabilities.
Security fixes are coming November 11 for Node.js 26.x, 24.x, and 22.x.
All three release lines are affected. Production users should take note.
Details: nodejs.org/en/blog/vulnerabi…
My favorite Musecase - find one event/outing to go every day
My favorite @Muse use case — asking it to mute my laptop while I’m in bed so the Codexes doing iOS simulator mic tests don’t wake me up 🥳
Maybe I should connect Muse to my laptop with Tailscale so it can control OpenCode 🤔
Having a ton of fun with this so far. Immediate reactions:
- The browser use on Muse is insanely fast. I haven't seen browser use this good yet
- Model has gotten a lot better over past week or so. Much less verbose. Feels natural bc it separates it's messages into multiply responses, emoji reacts to your messages, etc
- I love how you can ask it to create a custom avatar and it types as it works :)
There's a popular conception that a vegetarian diet cannot achieve healthy intake of protein. I spent some free time learning how to cook and what veg portion sizes can meet any protein/calorie ratio goal.
shukant.com/posts/a-full-spe…
One of the more surprising learnings is that staples like rice and wheat actually do have protein to meet USDA recommended daily allowances if you are at maintenance. But they are not protein dense enough to synthesize muscle or if you are cutting calories and want to maintain muscle mass.
Shukant Pal retweeted
Today we're launching Proliferate, open source Codex for working with any agent- here's how we use Proliferate to build Proliferate.
Lord, never click a large Excel sheet in Codex. It will permanently freeze Codex even if you quit and restart.
Shukant Pal retweeted
Why AI agents failed to get more adoption is really simple:
• They’re hard to create
• No one has time for it
• When they do, the agents suck
• And they break and are impossible to fix
In the end, doing the thing has been faster than automating it.
You solve this by making agents as easy to “build” as just Slacking a teammate.
Cutting the friction so much that your team starts building agents all day without even noticing.
Try today: lindy.ai












