Did you know that we have LinkedIn, Instagram and X accounts
that you can follow?
In this issue:
🤿Deep Dive: TypeSafe is building models for software
🤝From Our Friends: AI is saving you time. Your desk is taking it back.
🤖AI Made This: Interesting and inspiring creations made with AI
⚡Quick Hits: The latest in AI, tech, and productivity
👀On Our Radar: Something worth checking out before you go
⚒Tool Snapshots: Tools for AI, no-code, and productivity
🤿 DEEP DIVE
AI Built for Other Software
TypeSafe thinks people shouldn’t be AI’s only consumers.
Most frontier AI is pretty good at talking to people.
TypeSafe is asking a different question. What happens when the thing consuming the intelligence isn’t a person at all, but another piece of software?
That’s basically the idea behind the AI lab, which just emerged from stealth with $40 million in seed funding led by DCVC. It was founded by Diogo Almeida, a former OpenAI researcher and co-inventor of RLHF/ChatGPT, alongside Erik Gafni and Sasha Sheng.
But the funding isn’t really the interesting bit here.
It’s what they’re trying to build.
The chatbot qualities you might not want
TypeSafe’s argument is that the things that make an AI assistant useful to a human don’t necessarily make it useful inside production software.
Models can hallucinate. They can change their methodology between requests. They can introduce variability.
That might be manageable when there’s a person sitting there reading the answer. It’s a different proposition when software itself needs to use that intelligence to make decisions.
So TypeSafe is going after what it calls “machine-native AI.” Instead of intelligence primarily designed to interact with us, it wants intelligence that can behave more like a composable software primitive. Basically, something developers can reliably test, constrain, combine with other systems, and build on top of.
And I think that distinction is the whole story.
Almeida puts it pretty neatly. Most intelligence, in TypeSafe’s view, should eventually live inside software and run quietly in the background.
Jev shows what that looks like
Its first model is called Jev, a nod to Jevons Paradox.
According to TypeSafe, Jev delivers frontier-level intelligence with less than 100 milliseconds of latency and is up to 100x faster and less expensive than other frontier models.
There are two other details that make the broader idea easier to see.
Jev can process hundreds of outputs in parallel from one prompt. And it provides calibrated confidence scores, so developers can decide when their software should act autonomously and when it should defer.
That’s a pretty different framing from “ask the AI something and get an answer.”
The model is still in early access for select developers, though, with access currently waitlisted.
So for now, the bigger thing to watch is the idea itself. TypeSafe isn’t trying to make AI better at talking to us. It’s trying to make intelligence predictable, cheap, and fast enough that software can start using it as another building block.
The quick take
TypeSafe emerged from stealth with $40 million in seed funding led by DCVC.
Its focus is “machine-native AI” designed to work as a primitive inside software.
Its first model, Jev, runs at under 100 milliseconds of latency and, according to TypeSafe, is up to 100x faster and less expensive than other frontier models.
Jev supports hundreds of parallel outputs and calibrated confidence scores, but access is currently limited to select developers.
🤝 FROM OUR FRIENDS
You're reading this hunched
forward. You know you are.
I refuse to sit quietly in the background while you take damage from a chair.
Your physical therapist's future boat is being funded by this timeline. Ignore me and my tone sharpens.
Pick from 7 kinds of physical self neglect you need help with. Free download for Mac and Windows.
🤖 AI MADE THIS
Interesting and inspiring creations made with AI
Orbital Garden - an AI-built particle artwork you can actually play with
Orbital Garden is an interactive particle artwork with 48,000 points of light that shift between a flower, a gravity ring and a galaxy, and you can poke around with the whole thing yourself. I do like that the AI was used across the build, from the idea and maths to the code and interface, so it ends up feeling more like a small digital artwork than an AI demo.
AI used
GPT-6 Astra Ultra, according to the creator's documented creation record. The creator describes it as an iterative collaboration rather than a one-shot generation.
⚡ QUICK HITS
The latest in AI, tech, and productivity worth knowing
Meta is putting more features behind subscriptions. Meta One bundles extra AI usage and expression features across Instagram, Facebook, WhatsApp, and Meta AI, while Meta says the core experience will stay free. More than 50 features are rolling out, with individual bundles starting at $7.99 a month and creator and business bundles at $14.99.
Claude is getting rid of some tab confusion. Anthropic is bringing Claude chat, Cowork, and Artifacts into one interface, with Claude automatically routing requests instead of making people figure out which tab to use. It’s also adding dedicated tools for creating, editing, sharing, and exporting documents and presentations, with Pro and Max users getting them first.
👀 ON OUR RADAR
The best influencer marketing advice never comes from a report
It comes from someone who’s been there, done that, and has the receipts. Get ready for Return on Influence Festival ‘26. An entire day dedicated to influencer marketing with speakers running some of the best programs in the world.
October 21, 2026, 10 am - 5 pm ET. Free and online.
⚒ TOOL SNAPSHOTS
Futuristic tools within AI, no-code, and productivity
🎙️ Voiskey
Say the rough version and get something you can actually send.
Why it’s useful: Handy when typing feels too slow but you still want the message cleaned up for whoever’s reading it, without losing how you normally sound.
🧠 Thread
Gives all those random notes and thoughts somewhere to connect.
Why it’s useful: Useful if you capture a lot but rarely go back through it, since Thread can surface related memories and patterns instead of leaving everything buried in separate notes.
Lets your coding agents use whichever model makes sense at the time.
Why it’s useful: Useful if you work across Claude Code and Codex and want requests routed based on available quota and cost, without constantly switching between them yourself.
ℹ️ ABOUT US
The Intelligent Worker helps you to be more productive at work with AI, automation, no-code, and other technologies.
We like real, practical, and tangible use-cases and hate hand-wavy, theoretical, and abstract concepts that don’t drive real-world outcomes.
Our mission is to empower individuals, boost their productivity, and future-proof their careers.
We read all your comments - please provide your feedback!
Did you like today's email?
What more do you want to see in this newsletter?
