gorbypark

It depends if people care about it or not. I never once thought about accessibility in my non-AI authored app. I tried using the screen reader accessibility option and it was unusable.

I prompted Claude to "make my app accessible and usable with a screen reader" and it pretty much did a perfect job making it usable. *

* I'm not a a11y expert so "perfect job" might be an overstatement, but it made the app completely navigable by me using a screen reader.

Wow, odd. I haven't even set foot in Canada in 4.5 years and still have all the blocked features!

I am using OpenCode with the DeepSeek API with some pretty good results.

I have the complete opposite experience. Originally had a Canadian bought iPhone in Spain, had all the features a Canadian has and a European doesn’t (or vice versa). Upgraded to a Spanish bought iPhone and I am still a “Canadian”. I’ve been here for nearly 5 years but my Apple account is still fully Canadian (Canadian address, Canadian credit card on file). I think it’s Apple account location, maybe with some sort of system to allow people to switch countries but not allowing that to bypass restrictions? Or: that’s why a EU citizen can’t just switch their account location to unlock features?

But does this act like a debit account?

Yeah, it's a debit account. I'm in Spain and use Bizum frequently. It's just a "pay from your bank account" system.

You mostly type in your phone number, get a notification (or text, depends on the bank) and open up your banking app and approve the transaction.

You can send money person to person as well.

Many European countries have a similar system. Wero is "just" stitching the national systems together into a EU wide one.

Credit cards with rewards and points are pretty rare in Europe, and if they do exist, pale in comparison to what you can get in the US/Canada. It depends on how you look at it, but it's kinda good. The EU caps credit card transaction fees at 0.3% and debit transactions at 0.2% iirc, versus in the US/Canada where they are frequently 2-3%. In theory this cost is just passed onto the consumer, so paying an extra 2-3% to get 2-3% back in points or whatever.

I saw a video from the UK government with a guy walking down the aisle of a Lidl talking about it! I quickly went to Lidl.es to look as these 800w balcony solar setups are legal here in Spain. Unfortunately they don’t sell the here! Maybe once the UK does it they’ll start.

The micro inverter (most of these balcony kits use micro inverters) uses the grid as the reference. Most of these inverters will actually do nothing when the grid goes down. Like, they shut down for safety so you are not back feeding the grid, but even if you had some sort of back feeding isolation going on, they would still do nothing because they don’t have the reference of the grid.

It’s a downside of many grid tied residential systems (even large ones). No grid = no solar.

The Enphase IQ8 series is one of the first mass market micro inverters based systems to have the ability to make its own tiny electric island when the grid goes down. Requires an isolation switch and a relatively power hungry controller to use that feature, though. I looked into them for a balcony solar setup but it would be way overkill to run a full on controller for 800w of solar!

The best way for a small setup is just have a small “solar generator” battery that can take MC4 connectors as input. Prolonged power outage? Unplug the inverter, plug in the battery.

I’m going to buck the trend and say it’s really not that complex. AFAIK they are using Ink, which is React with a TUI renderer.

Cue I could build it in a weekend vibes, I built my own agent TUI using the OpenAI agent SDK and Ink. Of course it’s not as fleshed out as Claude, but it supports git work trees for multi agent, slash commands, human in the loop prompts and etc. If I point it at the Anthropic models it more or less produces results as m good as the real Claude TUI.

I actually “decompiled” the Claude tools and prompts and recreated them. As of 6 months ago Claude was 15 tools, mostly pretty basic (list for, read file, wrote file, bash, etc) with some very clever prompts, especially the task tool it uses to do the quasi planning mode task bullets (even when not in planning mode).

Honestly the idea of bringing this all together with an affordable monthly service and obviously some seriously creative “prompt engineers” is the magic/hard part (and making the model itself, obviously).

I have the Polymarket app installed because it's a React Native app and I installed a bunch of the top apps using RN as research. About a week ago I got a notification that "Polymarket is now legal in the US!". I'm not from the US so not sure why I got the notification. Anyways, seems like something has changed.

Does it work in dark mode? I guess if it does it would have to make the background apps brighter?

This is correct more or less. The Ikea hub has had the ability to bridge its zigbee devices to Matter for a while now. So in my case, Apple Home has no idea my lights and switches are not Matter, they just show up there even though they are actually zigbee.

Ikea recently did an update to enable the hub to be a Matter controller itself (over thread or Wifi). This means you can add matter devices to the Ikea hub directly and use the Ikea Home Smart app the control them instead of Apple Home or etc. You can add non-Ikea matter devices as well as Ikea matter devices (when they are released).

Yes and no. Zigbee is both the transport and the protocol, whereas Matter is a protocol but can run over different transports. Most common in Thread or WiFi, but it could be over ethernet or anything else, really. I would say Matter is not derived from zigbee, but Thread could be considered a derivative.

This seems to be a Python based Jupiter notebook (style?) thing for collaboratively working with GIS data/visualizations.

OpenStreetMap is a project to "map the world". In the end, OpenStreetMap provides data (and map tiles) for other things to use.

Going out on a limb (since I haven't used it) but JupyterGIS can probably make use of OSM data, along with other data sources.

Jeep is horrible. I was gifted a 2007 Jeep Commander, which was Jeep's "answer" to the Hummer. This was in like 2017, so it was 10 years old at that point. Anyways, it wouldn't shift into 4x4 mode, and after some internet sleuthing I found out there was a (now second) firmware update the dealership could do to hopefully fix the issue. I don't remember the exact details, but basically there was a hardware flaw in the module controlling the transfer case, and when it failed the vehicle would go into neutral, which obviously could be quite dangerous depending on where you were parked / what you were doing.

Instead of fixing the actual hardware issue, they did a recall that was some sort of black magic with a firmware update to "fix" the issue. According to the internet, this fix temporarily worked, with pretty much all of them failing again, conveniently after the vehicle was out of warranty.

Anyways, there was a second firmware update, that I had done 10 years after the vehicle was made, that more or less actually "fixed" the issue. Apparently the issue (according to Jeep forums, so take with a grain of salt) was due to some traces being undersized on the PCB, so the fix was to drop the voltage and/or current being sent, and then more or less disabling the safety sensors that would complain about low voltage. After the second firmware update, it would shift into 4x4 about 1 out of 4 attempts (otherwise just failing with "couldn't shift into 4x4" on the screen), and that was the final thing that could be done.

It took Jeep about 4 or 5 years to issue that final firmware update, probably to try and avoid a class action lawsuit over 90% of the vehicles 4x4 system failing just outside of the warranty period!

You can, but the issue is usability..when I'm watching TV, I want to just be able to flip open an app and say "I'm in London!" and watch BBC, then the same for Canada and etc. I don't want to be fiddling with a VPN and switching routes on some separate device / switching the entire wifi network or etc.

I am surprised to see three react native focused companies on the list. Expo, Software Mansion and Callstack are by far the big dogs in the RN ecosystem.

I only ever use a VPN to access region blocked content and the occasional "linux iso" torrent..I tried Mullvad first, but they just don't play the game of cat and mouse with the streaming providers and all their IPs are pretty much blocked. I have about a 95% success rate with NordVPN (except for Amazon Prime video which have some sort of wizardry and always are able to detect VPNs).

It's a shame because Mullvad has a deal with Tailscale where you can sign up for Mullvad through Tailscale and use any of their servers as a Tailscale exit node. It's super slick and nice since Tailscale has really decent apps for nearly everything (even Apple TV, etc) and I already have a decently sized Tailnet of all my devices / ssh accessible things.

That's a lot. I always had this idea in the back of my mind that British Columbia should get in on the AI game and try and get data centers located in BC because we generally have a lot of "excess" hydro generation capacity. There's a new mega dam recently opened that had lots of criticism about it being "unneeded".

That mega dam (Site C) produces 1.1GW of energy.

What keeps me and a lot of people/companies on React is React Native (and React Native web / strict dom). I'm sure we could move over to Svelt or Vue or any other number of frameworks on the web, however having a shared codebase and/or shared components across native and web is a game changer and not currently possible with anything but React.

Especially since the $142 price doesn't even include the memory chips! The YouTuber had access to defective donor cards to pull memory modules from, so they hand wave that into being free.

In the end it's the same thing, but in many countries where iPhones are popular, it's more of the "anti status symbol" effect happening. An iPhone is not a status symbol anymore per se, however NOT having one is the thing that gives you a "lower" status.

This would be the way to go if on the flip side any part that was not iCloud locked could be paired without hassles. Phone stolen/lost/etc? Parts unusable. Phone iCloud unlocked? Parts free for use. Of course this depends on mitigating various ways thieves can iCloud unlock stolen phones..I think the current method is snatching the phone while it's in use and iCloud unlocking it? However that doesn't make much sense since I assume you need some sort of password to do so even if the phone is physically unlocked?

React native support for SPM is non-existent for the most part. There's some work on porting to SPM but it's gonna be a while before anything becomes stable. It's going to break a huge amount of 3rd party packages. A bit of shame this wasn't done earlier as the RN ecosystem just went (or, is still going) through a migration to the "new architecture" that required most 3rd party packages that use native code to be "ported" over. Could have been a two-for-one kinda thing!

Room temperature super conductors would be huge for the EV charging space. One big thing holding back EV transport trucks is charging capacity. The infrastructure to support charging a fleet of EV trucks at 1MW+ each is insane with current tech.

Even supporting a bank of 350KW chargers is pretty nuts like you see in some highway rest stops (or Tesla Supercharger stations). Locations are limited by the proximity to high voltage transmission lines right now (ie: it's really expensive to push that much current for any significant distance).

If we had cheap and ubiquitous super conductors that could be run like regular ole medium voltage electrical cables....game changer.

You could try sandbox-exec. It’s kind of depreciated but was more or less designed for this exact use case I think. It’s too bad Apple doesn’t really support it anymore (although it still works in my limited testing!)

For sure, Claude Code specifically has a grep and glob tool (as well as access to bash). The LLM will decide to use those tools to find files, then once it has found matches to what it's looking for, it reads those files into its context. It will then decide if it indeed was a successful match and start working on that file. There's a lot of times it finds a match using grep, reads that into its context and then decides "no, that's not what I was actually looking for..." and moves on.

Unlike some other AI coding IDEs, Claude Code doesn't either keep your entire codebase in context, or use some sort of vector representation and vector search of your code. It basically uses grep to find a keyword it's looking for then reads the file (it can also just read a few lines of a file too) into its context.

It seems to be the "magic" of Claude Code..it's very UNIX like, uses existing tools to do what it needs to do instead of trying to re-invent the wheel, so to speak.

These being mixture of expert (MOE) models should help. The 20b model only has 3.6b params active at any one time, so minus a bit of overhead the speed should be like running a 3.6b model (while still requiring the RAM of a 20b model).

Here's the ollama version (4.6bit quant, I think?) run with --verbose total duration: 21.193519667s load duration: 94.88375ms prompt eval count: 77 token(s) prompt eval duration: 1.482405875s prompt eval rate: 51.94 tokens/s eval count: 308 token(s) eval duration: 19.615023208s eval rate: 15.70 tokens/s

15 tokens/s is pretty decent for a low end MacBook Air (M2, 24gb of ram). Yes, it's not the ~250 tokens/s of 2.5-flash, but for my use case anything above 10 tokens/sec is good enough.

For sure, it was a bit of a simplified example, but you can imagine prompts that are asking the LLM to touch dozens of files at once. A lot of people will use fairly high level requests, "make a screen in my app that does this" and that ends up making screens, components, queries, local state changes, etc etc. It adds up quick having so much in the context at once (and if you /clear the context and ask it to do anything more than one off changes in a file, it's gonna grep and glob it's way through your code looking for the right files)

As a $20 month user, I can tell you in my experience it's "refactoring" jobs that really smash through those tokens quickly. If you do a "write a component that does this" kinda thing, you can use the $20 plan almost an unlimited amount of time. If you are doing "find all instances of ComponentFoo, change to ComponentBar, refactor each screen for correct usage of ComponentBar" kinda things, it's going to grep through your code, find multiple files, read all of them into context and start making changes one by one and/or spin up a subagent to do it. You'll be rate limited pretty quick doing things that way.

Is there really an off peak time, though? I think Anthropic is running on AWS with the big investment from Amazon, right? I'm sure there's some peaks and valleys but with the Americas, Europe and Asia being in different time zones I'd expect there'd be a somewhat "baseline" usage with peaks where the timezones overlap (European afternoons and American mornings, for example). I know in my case I get the most 503 overloaded errors in the European afternoon.