←
Michael Doise12,023 posts
Michael Doise avatar

Michael Doise

@mikedoise

iOS dev | Built Perspective Intelligence (on-device AI), Beyond The Gallery & more at @TechopolisApps | Tech, sci-fi, gaming | Making iOS accessible to everyone

◷ Joined March 2008▧ 1,278 Photos and videos
Michael Doise avatar
Michael Doise@mikedoise·Jun 14

Hello everyone. I am very excired to announce AFM Studio, the best way to test AI models and providers on macOS and iOS. The project is open source and contributions are welcome and encouraged. You can learn more and try it out by visiting the repo at github.com/Techopolis/AFM-St… We have support for the system model, PCC, and Core AI models currently. This even includes our own model registry that will be updated often. Enjoy! #buildInPublic, #OpenSource, #AppleAI

Michael Doise avatar
Michael Doise@mikedoise·2m

I absolutely love that I can create a mini-sidebar on the iPhone Duo and iPad messages app. That is so nice making that smaller to show more content. I wonder if that works on the Mac as well.

clem 🤗 avatar
⟳ Michael Doise retweeted
clem 🤗@ClementDelangue·21h

Introducing Qwen3.8-27B-pi-adapted 🤗 We adapted Qwen's model specifically for the pi coding agent using reinforcement learning and a LoRA adapter. The results, measured as pass@1: 🚀 SWE-bench Pro (Hard 51): 29.4% → 62.7% 💻 Terminal-Bench 2.1: 39.5% → 55.8% ⚡ 55–62% fewer tool calls per task This is the power of open models: you can adapt them to YOUR tools and YOUR use cases. Let’s build! 💛 Open weights: huggingface.co/harness-hub/Q… Methodology: huggingface.co/blog/harness-…

Post attachment
Michael Doise avatar
Michael Doise@mikedoise·22h

Wired needs to work on its article titles. The iPhone Duo did NOT get a limited version of Face ID. The ability to unlock with the front camera does not at all use Face ID technology.

Michael Doise avatar
Michael Doise@mikedoise·23h

Looks like you will need an iCloud+ 200 GB storage plan to use Apple Foundation Models Pro in the Shortcuts app. 2 TB gives you 5x usage.

Michael Doise avatar
Michael Doise@mikedoise·23h

I’m really excited to give GPT 6.2 Sol and Astra a try today. I’ve really liked Codex and my Dot, so let’s see what’s possible!

Michael Doise avatar
Michael Doise@mikedoise·Nov 16

What?! The iPhone Duo get’s a new face unlock feature? I didn’t see that coming!

Tom Turney avatar
⟳ Michael Doise retweeted
Tom Turney@no_stp_on_snek·Nov 16

It begins

📢 Meet Qwen4-Max, your new AI teammate for work, creativity, and discovery! Qwen4-Max is built on a new MoE architecture with native vision and a 1M-token context window. Highlights: - Outperforms Qwen3.8-Max with up to ~50% fewer tokens, while average cost per task drops over 60% in our internal evaluation. - Advances real-world productivity through greater proficiency with professional tools, stronger long-horizon execution, and higher-quality deliverables. - Handles real tasks, including 3D modeling, game development, manufacturing, finance, law, healthcare, and frontier scientific discovery. - $1.5/M tokens for input, $4.5/M tokens for output, and $0.075/M tokens for implicit caching — all lower than Qwen3.8-Max, with implicit caching down by 70%. The API is live on QwenCloud. You can also try Qwen4-Max on Qwen Studio, QwenWork and Qoder. Start building today! 🚀 - Blog: qwen.ai/blog?id=qwen4-max - API: qwencloud.com/models/qwen4-m… - Token Plan: qwencloud.com/pricing/token-… More to come! The open weights of smaller Qwen4 models will be released. Stay tuned. 👀

Post attachment
Michael Doise avatar
Michael Doise@mikedoise·Nov 15

Ok so I am going to weigh in here and say the I would spend 30k on a cyber cab even if I could only use it here in Austin. I live here, so that box is checked. I still don't think they would sell one to a legally blind guy though. @Tesla If that isn't an issue, then let me know, because that would be amazing and I'd do it!

Michael Doise avatar
Michael Doise@mikedoise·Nov 15

Don't learn to code to get things done in 2026. Learn to code because it is fascinating, and so you can direct your AI agent to be more productive. Telling your app that the code doesn't work is not as effective as saying that it needs to try a specific coding pattern or library that it may not have thought of.

Michael Doise avatar
Michael Doise@mikedoise·Nov 15

Ok Here is what I think we are going to do. I've been testing our tool routing for a while now, and I think we've come up with a good plan for Perspective Intelligence. LFM3 1.7B will be the recommendation for regular chat on iPhones. You can then pick Gemma 4 E2B or E4B QAT if you don't mind the speed hit. AFM will be an option, but it will not work as well as the other models. Agents will recommend Gemma 4 E2B. I think we are going to stick with these four models on iPhone for now. Mac users can use more models. Our vision work will be completed by Apple's models.

Michael Doise avatar
Michael Doise@mikedoise·Nov 15

I used Qwen 3.8 27B running through OpenCode and TensorFold to start learning the basics of Rust yesterday. It was completely free learning. How can you beat that! On-device AI is the way to go!

Michael Doise avatar
Michael Doise@mikedoise·Nov 14

The main barrier to entry for on-device AI is that it is hard to understand if you are a standard technology user. We should all work to make it just as easy for everyone as cloud AI is now.