I was so excited when they presented "improved dictation" during WWDC, and people reviewing the developer previews seem to think it was a stand-in replacement for things like Whispr Flow, but I am little disappointed with the eventual release. It's not a disaster, but I feel like there's so many small little things they could add to meaningfully improve it. Things like adding a dictionary of words that you can maintain for proper nouns or choosing which mic to use for dictation.
Interestingly enough, I was working on something very similar this past few of days[1]! A couple of things that I did differently -
One, in the post-transcription cleanup phase, instead of programmatically fixing the text, I just feed the text with a prompt + dictionary of words into an LLM. Initially tried doing this with the on-device model from Apple in macOS 27, but that was too slow. I've now been playing around with gpt-oss-20b on Groq, which appears to be fast and effective. Although I'm curious how well your programmatic approach works, I might try some sort of a bake off over the next few days. But in the long run, I think I might still stick with an LLM-based formatter because you can have far more interesting formatting considerations or formatting prompts based on the app that you are using or the style or tone of your dictated text.
Two - When building the dictionary of words to fix, I'm also including a recording of the word and having the transcription model just predict draft out variants to use as replacements.
The Netherlands is preventing is trying to prevent hyperscale data centers from being built because they require as much energy is a year as a small city. I can't imagine dismissing the import of a small city's worth of Indians quarterly as trivial.
Outside of small bubbles, it’s hard to believe that anyone really cares, since it’s a resolved problem in most of the world. Without exaggeration, every person I’ve met from Europe, LATAM, East/West/South Asia, Africa, and Australia has by default asked me, “Do you use WhatsApp (or its local equivalent)?” when we decided to keep in touch.
I agree it has some monopolistic tendencies in the US and Canada and signals “iPhone is better!” when someone is what they call “a green bubble.” However, in the end, every group chat just gets created in an OS-agnostic app if one of the users has an Android. So, from my perspective, it’s a wasteful allocation of resources if you’re trying to “make it a good experience for the customers to communicate with non-customers” since it doesn’t really get “much better” than what we have now.
As a person who has friends with Androids, I have no idea about the video quality through MMS. We just use WhatsApp, and although it affects the quality, it’s generally good enough.
My point isn’t that this upgrade isn’t useful. I meant, why would a private company allocate resources to fix a problem that barely affects their users?
> why would a private company allocate resources to fix a problem that barely affects their users
It doesn’t barely affect users. That’s the point. In the US 50% of people use android. In the rest of the world that numbers a lot higher. Anything that goes over MMS sucks. Having a single android person in a group chat means falling back to SMS/MMS and everything sucks.
RCS is a huge quality of life improvement when texting non-iPhones for people who do that on a regular basis.
Again, this argument ignores the existence of OS-agnostic apps like WhatsApp. Who uses MMS? I might be extremely wrong, as I couldn’t find numbers online. But just from personal experience as well, outside of US, WhatsApp and its alternatives are defacto standard over SMS/MMS. Significant chunk of countries offer free data for WhatsApp/Messenger usage, and SMS/MMS is still paid per text.
People are not going to switch to using SMS just because iPhone added RCS either.
There is something to say about the platform setting the bar too low.
You're right that people move to third party clients, but the issue with Apple/Google screwing messaging is the third party services become de-facto while being sub-par.
In particular rival services don't get privileged access, which impacts either service quality (in particular conditions to receive a message) or user QOL (battery drain, core featuees implementation lag etc.). Line is a prime example of that, with a crazy wide audience in specific countries yet it became a pile of turd software wise, being optimal on neither platform.
The default messaging app being somewhat interoperable is IMHO a huge deal.
I’m very curious about the numbers that Apple/Alphabet see internally, as I couldn’t find anything from quick searches. It’s just very hard to imagine that it’s an actual widespread issue. I can think of two demographics who experience this problem. First is older people (70+) who have never been exposed to apps, have an Android and text people on iOS. And the small groups of people that avoid “X, Y, Z” apps for ideological reasons, are on Android, but keep in touch with iOS users.
Normal people don’t care about open standards. They only care about outcomes.
I know people who have wondered why texting to/from Android phones “doesn’t work” since pictures/video look terrible. This is a problem for their customers.
It wouldn’t surprise me if they would’ve done this without regulatory pressure, just perhaps not as soon.
There are plenty of small cars made outside of China. Why do you think not a single auto manufacturer worldwide bothers to export their best selling models to the US? The market here wants SUVs and trucks, not $15K subcompacts.
Outside of Chinese-made vehicles the US has probably the most open auto import market in the world. Most manufacturers enjoy a 2.5% tariff, which is a joke, and not reciprocated anywhere in the world. All these foreign manufacturers are selling $25-50K+ cars just fine (over 4 million of them every year in fact). It's the US market that is rejecting lower priced cars.
higher sounds like a quality argument, which may be true. they are not stricter though, which is what they'd need to be to refute the parents point. take headlights, for example, where the us's requirements are significantly stricter than european ones.
You might be able to run on a 5W supply, but my experience is that booting on a less than 15W supply is a recipe for corrupted SD cards and hard-to-trace bugs.
My biggest frustration with hobbyist boards is how little power the can source to e.g. the USB ports. Its super frustrating to need a powered USB hub to add e.g. a USB indicator light. The fact that you can power the RPi from the powered USB hub seems hacky, but maybe that was always the intent.
Try to run even a Pi 3 on 5W and let me know how it goes. You'll get undervolt warnings like crazy. Not sure if you looked at the article but these are the power adapters listed. These SBCs are not exactly low power anymore:
You'd need to limit the CPU operating points or offline most of the CPU cores to get lower max power consumption under load. 5W is too strict for Orange Pi 5 Plus.
And of course forget nvme if you want predictable low power consumption. nvme alone is 5-10W load. The board itself can run under 5W, but you'd have to carefully consider what peripherals you can use and how.
M600 idles around 5-8 watts. If you spank it hard it gets up to 12W. That has a 500Gb SATA SSD and 8Gb of RAM and ethernet.
Note that the M600 does not have any active cooling. The entire unit is a big heatsink. The thing will run as a headless network computer with just the power supply (a proper Lenovo SMPS brick) and an ethernet port connected.
power usage on a computer is not constant by any measure.
power usage spikes with CPU activity (among many other things) and if your power supply can't deliver the current that is required, voltage will drop, and things start failing, but only for 1ms or maybe even less. maybe even a single clock cycle in some cases.
if you can't keep the board powered fully for every clock cycle, then you are going to have problems, and that's true for any computer, not just the Pi.
I felt like the system that’s in the portal from Facebook would be the ideal vc setup for my parents (i.e. video calling on a tv). Though, there’s no way I’d want to put a camera from Facebook in my parents house. I did try Google duo (I guess it’s call meet now?) with a Google tv, but since it was with a 3rd party web cam, things like autofocus didn’t work as well as I’d like.
You can always boot from your mbp? Treat it like an external bootable storage system. Yes you won’t be able to use the display on the laptop, but you will be able to run your applications on the iMac. (Assuming that matters more? For me it does. I can’t stand how slow docker is on my mbp, pretty much puts my cpu to 100% and has the fans running at full speed)
I wanted this so much. Turns out that doesn't work with encrypted drives which is mandatory. Target disk mode only allows booting in non encrypted drives or perhaps non t2 MacBooks.
