Doing iconic shit DevRel at @TypeSafeAI Queer as fuck
Joined June 2025
- Tweets298
- Following63
- Followers2K
- Likes297
When I tell you I am obsessed with getting my custom Alexa-replacement voice assistant to have as little latency as possible, I want you to believe me
Constantly pouring over timing traces to figure out where the inefficiencies are
Pondering on this more, I think the cornerstone of what makes personal software so possible is not that it's become so easy to build, but rather it's about what you DON'T have to build.
Personal software only needs to function on your device. With your data, and your processes, and your OS version and your desires.
You don't have to create a new user experience, you don't have to handle accounts, you don't have to worry about handling others' data, you don't have to handle backwards compatibility, you don't have to do migrations, you don't have to worry about uptime.
You don't even need a preferences dialogue!
Those of us that have built Software Products are all too familiar with THAT stuff being what takes like 90% of the effort. The core product functionality is (often) the easy and fun part, and all of the other stuff is just what has to happen to make it be a product that others can use. But personal software doesn't even need to *dream* of being a product. It's exactly what you need it to be, and nothing more.
The era of personal software is so wild
The other day I was bemoaning having to start up some services in terminal windows each time I rebooted, and I was like hey why not just make an app that does exactly what I want it to do. By the end of the day I have Coven, a native MacOS app allowing me to manage little stacks of local services. It uses process-compose under the hood so it feels a lot like managing docker stacks, which I like
In fact it was this realization that has seriously accelerated my personal projects. I frequently catch myself starting to develop a feature that's really only needed for a multi-user environment. My sensibilities have always leaned towards doing this in the stuff I make, but freeing me from that has allowed me to make things so much more personalized
Having built many dozens of Jev-powered projects and demos over the past few months, I can say with confidence my #1 advice for how to use Jev is:
Just put Jev everywhere.
Every time I think of a new (not Jev specific) project, I think to myself "what could Jev do in this system to make it cooler/smarter/weirder/sillier/whateverer" and then I do it.
The objective is not to use it in places where it makes total sense. The objective is to continue to expand my mind to see Jev the way our long term vision has always been: as an enabler of a new kind of smart software. The more you train your brain to truly think of Jev as smart primitives that can slot in to code in anywhere a primitive could live, the more you find novel use-cases.
Don't stop building. Don't stop playing. Experiment forever and ever and have fun with it!
When I made Agent Tower, it didn't use Jev.
But then I thought "the decorations in the room could be based on the vibes of the thread, which Jev could determine!"
And then later I thought "I could use Jev to shorten thread titles to just the words that fit!"
And then "the subagents could have different outfits based on what they're being asked to do, like artists and coders and reviewers, which Jev can determine!"
holy shit MacOS 27 finally added a way to see the menubar icons that overflow in to the notch.
It took literally FIVE YEARS for them to address this. Previously any icons that fell under the notch were invisible, forever, with no way to access them natively. Bonkers
The era of personal software is so wild
The other day I was bemoaning having to start up some services in terminal windows each time I rebooted, and I was like hey why not just make an app that does exactly what I want it to do. By the end of the day I have Coven, a native MacOS app allowing me to manage little stacks of local services. It uses process-compose under the hood so it feels a lot like managing docker stacks, which I like
Honestly, respect. I've long felt that desk workers should be provided with exactly whatever equipment they prefer at their workstation. They're sitting there and using those tools all day long every single day. Even a 1% reduction in productivity for a tech worker costs way more than whatever laptop or monitor or fancy keyboard they like using
This quoted post is unavailable.
Waking up to design changes in @t3dotcodes:nightly every morning is how I know I'm still alive
Are you at @githubuniverse this week? Be sure to catch my session "Jev: Shipping a category-creator" where I'll talk all about the journey of bringing Jev to market from the DevRel perspective.
Thursday @ 1:15pm!
reg.githubuniverse.com/flow/…
One of my favorite Jev patterns is "line-by-line search", where you put an entire document in state, and ask a question for each line (or paragraph). This is super useful when you want to find the most relevant line for a query.
Use nouls when you expect one dominate answer, use scores when you want to highlight lines by overall relevance. The former is good for answering a specific question while the latter is good for finding all relevant passages.
If you want to try this in your own project, point your agent (or your own eyeballs, if you're old fashioned like that) at this cookbook! docs.typesafe.ai/cookbooks/s…
NetworkChuck built his own version of my Intelliprompter and shows it off in this video!
This was such a fun conversation. Thanks for the great 0-1 explainer of what Jev is and how to use it, @NetworkChuck!
Allie the Icon retweeted
Thanks to @allietheicon and @CompleteSkeptic for the help with this video!!
@typesafeai jev is pretty sick.
Are you in SF? Do you want to chat with yours truly about Jev? Join me and LangChain on November 5th for an evening all about adding a little judgment to your agentic workflows
Join LangChain + @typesafeAI on November 5th for an evening of talks on building with decision models and using them to evaluate agents.
Hear from @sydneyrunkle, Seán Roche, and @allietheicon.
Interested in attending? RSVP today! luma.com/iuc3ostf
Allie the Icon retweeted
We do AI video but I HATE this.
@Kurz_Gesagt was never about the visuals. I'm pretty sure most of that 1,200 hours goes to research and writing. The fact that thousands are now going to publish these kind of slop videos with a single Claude prompt will only make people click away from gems like Kurzgesagt because they will think it's AI too.
Same has happened to channels like Fern and @RealEngYT.
I'm also surprised that people I respect like @DeryaTR_ are cheering for it.
Kurzgesagt has 25M subs. Millions of views on EVERY single video 🤯
nobody copies them for a reason: one video takes 1,200 hours and a studio of 50+ animators.
we rebuilt their format with Opus 5.5 + @HyperFrames_ Studio. 100% code.
a full 11-minute explainer. made in ~2 hours. under $50.
want the template? we’re open-sourcing it for free 👇
This is siiiiiiick
Pre-launch I tried to build something like this but never got it to work or look half as good as this
It's SUCH a useful learning tool to see how Jev primitives are composed together in a system
Currently working on adding support for more complex Jev flows to the Jev Devtools 👀
This includes an API under jev-devtools/flows to connect Jev calls together (with meta data) which allows the DevTools to map out the diagrams for the flows to make it all easier to visualize.
All still very WIP. Not even sure where this is going anymore 🫠
And thanks @allietheicon for the idea! I hope this is kinda what you ment haha
