@heyjessharris It’s great that you can supplement with AI without letting it take over. I do think there’s still a need for a ‘voice’, but letting AI help critique and plan is game changing.
I'm reading more AI-written docs and messages than ever. A lot of them are hard to get through.
My advice has been: if you don't want to read it yourself, don't ask someone else to.
But then I read my own writing and realize it isn't always at the quality I'd like either.
So I'm trying to improve. Writing things myself still feels important.
What’s actually helped you become a better writer?
@hemal Congrats! Super exciting! Just enjoy it, you’ll get tonnes of advice on how to do this and that and what to worry about. But it’s such a fun time, make sure you take the moment to live it!
It’s also very cool we’re still using git and pushing fixes upstream. Agentic workflows are pushing the limits of what are our systems can do, but sharing those wins is so important.
Since this was turned on, I forgot all about it.
That’s a compliment.
Before; I was raging at git every day, fetching slow, pushing aggravating, merging impossible.
After; fetching never a problem, agents pushing all the time, merging simple. And I didn’t have to change my workflow.
It’s been so good, I forgot all about it.
We built Pando, a new Git backend for our monorepo, to handle more pushes and fetches without changing how our engineers and agents use Git.
Pando cut push latency by ~80% and now serves nearly a petabyte of Git data a day. We can grow Git capacity alongside agent usage.
The taste argument only exists in the tech bubble.
The rest of the world doesn't talk about whether taste is dead. They just want products that are simple, frictionless, and do the job.
And even when bots do the clicking, someone still has to design what they see, what they ask you to approve, and what they show you after.
Taste isn't going anywhere. It's just getting more users.
For years, accessibility was an afterthought for most teams.
Now it's more important than ever. Agents read your app the same way a screen reader does.
Web apps have always treated it as a secondary thing. Electron apps should think the hardest about this, since they're web UIs dressed up as desktop apps.
Apps agents can navigate will win. The rest get left behind.
Honestly, it's exciting. Even if teams come for the agents and not for accessibility itself, the people who rely on screen readers win too.
Introducing agent-desktop
Computer use for your agents
▲ E2E test desktop apps with agents
▲ Reads apps via the accessibility tree
▲ Drills into part of a window to save tokens
▲ Runs many agents on one app at once
▲ Works on macOS, Windows & Linux
npm i -g agent-desktop