Johannes Erhardt retweeted
Hey Swift community! Something exciting is happening.
Until now, AI coding tools barely managed Swift syntax. They’d choke on SwiftUI, Concurrency, or Hummingbird, and needed constant handholding.
That changed with Droid. @FactoryAI’s Droid just gets Swift.
Last weekend I started a project from scratch with a single prompt. No hand-crafted scaffolding. And it built a working Swift 6.2 concurrency package. An absolute first.
And it’s not just me. @rudrank opened a PR in MLX-Swift today that was 100 % coded by Droid.
My first Droid project was an experiment. One naive prompt, low expectations, and somehow it just worked.
So I’m doing a little experiment:
📘 Tonight I’ll RTFM
🚀 Tomorrow morning I’ll start a brand-new Droid Swift project — and live-tweet the whole thing. Let's see how far we can push this thing.
I’ll probably do a ton of things wrong (and hope @bentossell can rescue me if I get lost 😅).
Here's the plan:
If you’ve followed me, you know I’ve ranted for years about the lack of good Swift AI libraries. Even Apple’s own MLX-Swift trails far behind its Python equivalent in features and performance. It’s frustrating.
So I’ve got a long list of things to fix and create, starting with a performant TextView that supports both Markdown and LaTeX, and is optimized for streaming, and actually lets you select the whole message (not just one paragraph). That doesn’t exist in Swift today.
Tomorrow I'll start with this. It's not an easy project, and I’ll be impressed if it works first try.
If you’re missing any tools for AI in Swift, comment below and I’ll add them to the list. Let’s fix this ecosystem together. Swift deserves better. 💪
I just published my first project that was 100% coded by @FactoryAI’s Droid: a Swift SDK for OpenAI’s new Responses and Conversations API.
A few fun facts:
- Uses Swift 6.2 and full concurrency (something I’m convinced no human can do)
- I used GPT Codex as the Droid model
- Never experienced infinite loops, and every single build compiled on the first try — impressive, especially after battling Windsurf and Cursor.
- Droid chose an MVVM architecture on its own. It’s a bit over-engineered imo, so I might ask it to simplify things later. But it works, and that's great
- Includes a fully human-coded demo app supporting @openai, @groq, @lmstudio, and an internal @PicoGPT build (beta coming soon).
Johannes Erhardt retweeted
🇪🇺 EU Data Residency has landed in Notion.
Now you can keep your Enterprise workspace’s data stored in the EU to meet compliance + sovereignty requirements — without giving up any of Notion’s collaborative features 🌍 ✨
Your data stays local, your team stays connected.
Johannes Erhardt retweeted
New versions of Phoenix come with an AGENTS .md file that makes Claude/Cursor/etc work better with your codebase.
Elixir ahead of the curve as always:
Johannes Erhardt retweeted
Phoenix LiveView 1.1 is an exciting release in a lot of ways, but the highlight for me is the JS is getting types! Huge thanks to the maintainers 💜 #elixirlang
github.com/phoenixframework/…
Johannes Erhardt retweeted
#news Bear Web is now in public beta, and you are invited! 🚀
This is a big step toward making Bear available wherever you need it, and we’d love for you to give it a try.
Learn more and join Beta here:
community.bear.app/t/bear-we…
Johannes Erhardt retweeted
“This means it's finally possible to build and deploy iOS apps from Linux and Windows (WSL)” forums.swift.org/t/xtool-cro…
Johannes Erhardt retweeted
New feature this week in #ProtonMail for desktop!
🔄Now you can switch between your Proton accounts from the Proton Mail desktop app, and stay logged in to all of them.
Available for #Windows, #macOS, and #Linux: proton.me/mail/download
Johannes Erhardt retweeted
📣 MAX 24.4 ⚡️ is here! Featuring:
⚛️ New quantization APIs, with SOTA CPU performance for LLMs!
🍎 MAX for macOS!
🦙 INT4 implementation of Llama3/2!
➕ Performance and standard library improvements for Mojo 🔥, including 200+ community contributions.
modular.com/blog/max-24-4-in…
Johannes Erhardt retweeted
Are you using the @1Password SSH agent? 🤔
Find out how you can set it up to perform signed Git operations in Tower! 👇
git-tower.com/blog/1password…
Johannes Erhardt retweeted
I wrote a little piece on how you can train your first Machine Learning model, and run it inside your iOS app via CoreML. No DataScience or Machine Learning experience required
Johannes Erhardt retweeted
👏Goodbye, Google Search👏
Starting today, you can now set Perplexity as your default search engine in Arc browser...
Johannes Erhardt retweeted
Encrypted email service Proton Mail gets its own desktop app tcrn.ch/4ay8Opx
Johannes Erhardt retweeted
my 🇩🇪 startup wishlist:
- ✅ ESOP
- ✅ online incorporation
- ❌ remove notaries
- ❌ allow SAFEs
- ❌ 10M shares
- ❌ 0€ Stammkapital
- ❌ Inc. (no GmbH)
Germany has done it!
For years, 🇩🇪 has had the worst stock options policy. Yesterday, it passed a new legislation that catapults Germany into having one of the best employee ownership schemes, globally.
After 5 years of #NotOptional, change is coming!
cnbc.com/2023/11/17/germany-…
Johannes Erhardt retweeted
Sam Altman on why you shouldn’t track absolute user growth in the early days of a startup
“Nothing but a great product will save you; you can get everything else right and it still won’t work.”
He points out that almost all startup founders get the following wrong:
“It is more important to have a small number of users that love you than a lot of users that like you… Eventually what you want of course is a lot of users that really love your product, but that’s almost impossible to do.”
In practice, you have two choices:
Deep and Narrow: “You have a small number of users that really love you and then find out how to find more and more of those users and broaden the appeal of the product.”
Shallow and Wide: “You can have a lot of people that sort of use the product once or twice and kind of like it and try to figure out how to get them more engaged over time.”
“With high confidence, I can say that you want to start with a small number of users that really love you. Almost all great companies have products that start this way.”
He argues that a good indicator of users loving your product is retention and frequency of use:
“In fact, I think this is so important that you actually shouldn’t track absolute growth in number of users in the early days of a startup. You should just track how often they’re using it… That’s a good early indicator of users that love you—better still is them spontaneously telling their friends to buy your product.”
Follow @startuparchive_ for more tactical startup advice!
Johannes Erhardt retweeted
Two incredible new breakthroughs - the world’s fastest unified inference engine, and Mojo 🔥 a new programming language for all of AI. Visit Modular.com now to sign up!
Watch everything we just announced in 121 seconds!🚀
vimeo.com/822336980
