Children’s author, pattern linguist, slow technologist, senior editor @every. And newish dad. Building lively worlds, real and imagined.
Mt. Detroit
- Tweets 503
- Following 367
- Followers 11,965
- Likes 3,999
Pinned Tweet
My new one for @every going into the weekend, about two favorite topics: woodworking and UI design (feat. @joshpuckett and @Wattenberger) every.to/p/designing-with-ai…
4
6
87
38,110
Jack Cheng retweeted
How to use jev #1: As an embedder
I've been using Jev for all kinds of things. This morning I had a realization I kind of like:
Use it to make non-black-box embeddings.
Instead of an embedding model spitting out 1,536 numbers that mean nothing, you ask Jev questions about each document. The answers become the vector.
An email in Cora:
"I got charged twice this month, pls fix asap"
[is_customer, urgent, about_billing, needs_reply]
[1.0, 0.9, 1.0, 1.0]
A newsletter:
[0.0, 0.0, 0.0, 0.1]
A friend asking about lunch:
[0.0, 0.1, 0.0, 0.7]
Then it's just old-school cosine similarity search. Search "billing issues from customers" as [1, 0.5, 1, 0.5] and the double charge comes out on top.
Same idea for our articles at Every:
[is_tutorial, about_ai, contrarian, beginner_friendly]
Or support tickets:
[is_bug, angry, churn_risk, enterprise]
Every number has a name, so you can see why something matched. Need a new dimension? Add a question. Want urgent stuff first? Change the query vector.
Trying this in @CoraComputer now to make search fast.
24
48
861
185,456
Jack Cheng retweeted
DSPy 🤝 Jev
DSPy 3.4.0 was just released!
This release includes native support for Jev and System one models inside of DSPy! Use it with compatible signatures. This release also includes a brand new optimizer, ReAnchor, specifically for calibrating outputs with confidence.
6
30
1,741
Another experiment with using Jev on two ends of an email classification system (a Jev sandwich?) to figure out what’s worth my attention based on how well I slept and my current vibes
12
16
152
38,812
This was partly inspired by @danshipper’s thought experiment on “urgency” at the top his piece on seeing like a language model. Worth a read! every.to/chain-of-thought/se…
4
333
TODAY: Join our Write-along livestream at 1 p.m. ET featuring:
@kplikethebird's Compound Writing
@hammer_mt's writing process
@jackcheng's editing process
We'll be live on X shortly. See you soon and come with any questions!
2
3
14
3,007
the right way to use model capabilities is not to ship 10x more features to prod
it's to spend more time understanding your users, trying experiments, building prototypes, learning about things you don't understand so that you can ship things that actually work
275
615
8,218
382,640
Jack Cheng retweeted
my favorite thing i made with opus 5.5 is a personal origami tool that i can control with @claudeai
i can
- import designs from figma, web, or code
- vibe design like paper
- add patches and fine tune interactions
all while learning how it all works. its something ive always wanted
7
3
46
11,132
Jack Cheng retweeted
Agent nooks, infinite calendars, birthday candles, vibeometers - and emoji machines.
These are some of the device concepts we’ve been exploring for WORKS - our fast-to-market, limited-release promptable products, built in public and tested in the open. Each one offers a different way for agents to inhabit everyday objects.
In the last episode, we took you inside the Approach design studio. This time, @genmon and I tour the sprawling Figma board behind the process - and the fifty concepts the team worked up.
Throughout the design process, we’ve moved between methodical exploration and intuition. A framework helps you explore a direction. An odd reference sparks a conversation. Something catches your attention, and you follow it.
Where did we end up?
An agent with its own little nook on a shelf. A calendar that helps you see the flow of your day. An object that reflects the mood of a room back to you.
We can’t build them all. We need to pick five. It's a balance between what people might want and what we can actually make.
This is where we are with @inanimate_tech: sharing the possibilities while we’re still working out what they should become.
This is WORKS in Progress EP3: Exploring Our Maze of AI Hardware Concepts.
2
2
16
882
Jack Cheng retweeted
I'm still exploring what I can do with Jev but I made this little app which does the following
- talk about your day (or anything really)
- audio gets transcribed to text
- text is parsed for cues across 12 colours linked to an emotion [a logic I fed into this app for Jev to use]
- a painting is drawn with code in JavaScript using the colours Jev thinks suits the transcript cues well
- canvas has an underlying MIDI [utilises Tone.js]
- watch your words turn into a painting which turns into a melody.
App uses @typesafeai for Jev's colour + emotion inferences, @GeminiApp for audio transcription, @claudeai for the JavaScript painting code, @ChatGPT for front end UI refinement.
sayhue.vercel.app
16
17
246
10,966
Jack Cheng retweeted
My job at @every is basically benchmarks-as-a-service now: every.to/also-true-for-human…
11
8
143
11,398
I had the pleasure of editing this @every essay from Cyrus and Josh! Enlightening to see a professional creative studio test Astra’s vfx capabilities on their actual workflows every.to/vibe-check/vibe-che…
New essay for @every, by @afterimage
We believe we will look back on Astra as a turning point in how creatives use AI in video work.
For many months we were experimenting with different AI video approaches: generative workflows in ComfyUI, procedural animation with Python, motion design through the Cavalry and After Effects MCPs, and more.
All of these were fundamentally limited by a lack of control. Generative models could sometimes make cool images, but changing one thing without disturbing everything else was difficult.
Many of these generative tools weren't (and still aren't) built for true professional workflows.
On the other hand, general models starting around Opus 4.5 were decent at writing python code to deterministically edit images and videos, but the models were really bad at “seeing” the images and videos they were supposed to be editing, which made them impractical.
The day Astra launched, just from the rumors and marketing, our internal sense was "this might be the one". Then, when we saw the level of 3D detail people were able to achieve with Astra in Blender, we really suspected the model was finally able to "see" and iterate toward things that actually looked good.
While everyone raced to build cool (but often useless) demos of 3D worlds, we got interested in something else: compositing 3D assets into live action footage at commercial-grade quality. To this day, we are quite surprised by how few Astra demos focus on this ability. It feels like a breakthrough for us.
We tested a range of processes, from traditional VFX skills like masking and tracking to hybrid approaches like compositing Blender renders and generative imagery with real footage.
At the highest levels, Astra doesn’t replace someone with true domain expertise in VFX artistry. But, through these tests, it started to feel for the first time that the LLM was actually a viable tool in the kit. It worked well in concert with other tools to make professional results we could actually see ourselves using.
The coolest part about all this is that it makes it easy and painless to attempt visual effects on even the smallest projects and most off the cuff videos. We’re excited about the combination of casual footage and visual maximalism. Effects can now be completely over the top without feeling like the whole reason a shot exists.
Full "Vibe Check" with video examples & side by side Astra vs Sol comparisons (gift link) ➡️ every.to/vibe-check/vibe-che…
1
1
14
3,360
I actually don't disagree with the analysis in the meat of the video but:
a fake demo is different from a fake product
calling these fake demos is disingenuous and bait-y
a demo can be a tool for learning, exploring, and understanding – especially for people who don't call themselves developers
not everything needs to be productized
13
1
58
9,311
Jack Cheng retweeted
Since Jev launched Tuesday, there’s apparently a new thing at @every “Jev boys.”
Ever since the launch there’s been this flock of self-proclaimed “Jev boys” who carry Jev on hand at all times and constantly ask it what to do. They have their entire personality revolve around Jev, TypeSafe, and calibrated probabilities. When we went around talking about what people were trying this week, 4 or 5 of them said some variation of “I live by the Jev and die by the Jev.”
Just about an hour ago, I asked someone to do something and one of the Jev boys was bold enough to say “if Jev says I do it, otherwise I don’t.” I told him if he asked Jev instead of just doing the thing I was going to lose my mind.
He asked it anyway and it said to do it.
Thank god for that at least.
But then the other Jev boy asked the same question and Jev gave him a different answer.
He immediately pulled up the probability and started explaining that Jev was only 63% confident, so technically it was a judgment call.
I told him that yes. That is what being a person is.
Later, another Jev boy asked Jev whether he should go to a meeting. Jev said no.
He did not go to the meeting.
When someone asked where he was, he posted the Jev output in Slack.
At lunch, I watched two Jev boys ask Jev where they should eat. They got different answers and spent five minutes comparing confidence scores instead of picking somewhere.
One of them actually said “I’m going with Jev on this one.”
The other said “your Jev is wrong.”
I don’t even know what that means.
Now they’ve started asking Jev questions about each other.
Someone asked Jev if another person’s idea was good.
false
The person whose idea it was then asked Jev whether that was constructive feedback.
false
Neither of them has spoken since.
This afternoon one of the Jev boys told me he doesn’t make decisions anymore, he “routes them through Jev.”
I asked if he was joking.
He asked Jev.
It said false.
It has been three days since launch.
I really hope this dies off soon. I haven’t seen anything outside Every about “Jev boys,” so I’m guessing it’s just an Every thing.
Unless one of you has more information.
16
6
64
12,632
Another demo to show Jev’s speed. Here it takes all the english language posts from the Bluesky/AT protocol public firehose and 1) figures out which have an animal in them and 2) classifies whether it’s an actual animal, sports team, a figure of speech, etc, pretty much real-time
4
7
73
12,621
For more about how this model works, check out @hammer_mt’s vibe check every.to/also-true-for-human…
1
100
23,936