I also am looking forward to Apple smart glasses. I use Meta glasses because thre useful since I'm totally blind. I'd much rather have on device recognission though if possible. I also trust Apple more then Meta. At least I'm technically enclined enough to realize I shouldn't wear them in the bathroom or bedroom. At least when humans look at my AI prompts there maily seeing food boxes or computer bios screens.

Ever since Be My Eyes introduced there AI features it's my understanding that there's been a lot less need for volunteers. I'm totally blind, and started using the app once they added in AI. It works great for for things like reading food labels after my kids have movved things around, determining if the tv has been left on, etc. I think I would use the volunteer feature if I still lived alone, but I don't.

I haven’t had a single alert from it since the AI stuff rolled out

Thanks for creating this. I'm blind so the $16 USB tester off amazon to sort through my drawer of cables is not an option. This will stop me from needing to buy a sbc just so I have something running Linux to test cables.

How did Claude do at conforming to accessibility standards on macOS?

Curious whether you've tested the popover with VoiceOver yet. Menu bar SwiftUI apps usually nail the status item by default, but the popover content tends to be hit-or-miss depending on row layout. A stack of Text views walks cleanly, but custom-drawn rows often read as one blob. Worth confirming for jareds' use case.

What's the current situation for coding with Local LLM's on decent hardware? I have an M3 Max with 64 gb of ram and am thinking I should start looking at Ollama and Opencode? Is this a useful stack for smaller personal projects?

It’s getting there. You could give a try with qwen 3.6. It’s worth paying for better models in the cloud, but local models are now better than nothing.

One nice development recently was ollama's support for MLX optimization on Mac hardware. It's not obvious how to know you're using a model that works with it, yet, so it's rough around the edges.

https://ollama.com/blog/mlx

Use llama.cpp or better yet Unsloth Studio

I wasn't planning on buying one, but I'll add this to the list of app enabled coffee tech I refuse to buy. As someone who's blind I'm getting really tired of app enabled coffee equipment with no open source integrations or protocol documentation. Fellow also doesn't appear to make any effort to make there apps accessible. They have had there Aiden out for over a year and I still don't see any notes about accessibility in there app update. I'm not going to buy one and use the home assistant integration since that could break at any time. Luckily I'm more of a coffee drinker instead of espresso so the Ratio Four works well enough for single cups and half pots.

How long until the $10 Github Copilot subscription goes away? That was a great deal for my limited personal programming. The only reason I switched from it to Claude was to get coding and general ai in a single bill.

I think Github Copilot is in the process of slowly winding down right now. They've been putting very, very long (multiple day) rate limits on users for various esoteric reasons for weeks now and just yesterday or so paused signups.

It already did, you cannot sign up for it right now.

I just switched from the $10 Copilot subscription to a $20 Claude subscription to get general AI and coding in one bill. I guess I'll try out GPT Codex.

gpt allows you to wire their models into other CLI tools, I'm advising everyone I know to lean that direction. Not trying to become hostage to something like claude's ecosystem for the rest of my development career.

They will eventually converge --- it's only a matter of time.

All I want is a reasonably priced subscription combining both coding AI and general AI in a single bill for non professional use that allows me to opt out of my data being used for training. Google limits history to 72 hours if you opt out of training even if you pay them $20 a month which rules them out for me. I guess I'm going to try the $20 chat gpt plan. At this point I am wondering if I need to accept that were moving to a token based model and get comfortable with opencode and manually switching models.

Chatgpt $20 plan is a steal. There's nothing close.

Does anyone have experience with this Braille embosser? I haven't found any reviews or testimonials for this, but that may be because it appears a lot of the development has been done in France instead of the U.S. I'm totally blind and am interested in it for producing Braille graphics, specifically as a way to get an idea of what a design I create using OpenSCAD will look like before 3d printing it. I can't justify $2000+ for commercial embossers that can print graphics for what is a hobby I'm not sure if I will stick with.

The issue I have with using LLM's is the test code review. Often the LLM will make a 30 or 40 line change to the application code. I can easily review and comprehend this. Then I have to look at the 400 lines of generated test code. While it may be easy to understand there's a lot of it. Go through this cycle several times a day and I'm not convinced I'm doing a good review of the test code do to mental fatigue, who knows what I may be missing in the tests six hours into the work day?