I currently use MacWhisper and it is quite good, but it's great to see an alternative, especially as I've been looking to use more recent models!
I hope there will be a way to plug in other models: I currently work mostly with Whisper Large. Parakeet is slightly worse for non-English languages. But there are better recent developments.
I wish they had a "and we won't screw you in two weeks" plan at, say, 5x the price. It's worth it for my business, I'd pay it.
Should I switch back to API pricing? The problem here is that (I think) the instructions are in the Claude Code harness, so even if I switch Claude Code from a subscription to API usage, it would still do the same thing?
FWIW I've only ever been on the API based plan at work and we never seem to run into the majority of the problems people seem to be very vocal about. Outages still affect us, and we do have the intermittent voodoo feeling of "Claude seems stupider today", but nothing persistent.
Of course it's a stupid amount of money sometimes, but I generally feel like we get what we're paying for.
I never managed to get anything useful out of opencode, to be honest. I tried it many times, with various models. Claude Code always just worked better.
That's one of the possible explanations, but I think too many people are seeing the same symptoms (and some actually measured them).
An "economical explanation" is actually that Anthropic subscriptions are heavily subsidized and after a while they realized that they need to make Claude be more stingy with thinking tokens. So they modified the instructions and this is the result.
> but I think too many people are seeing the same symptoms (and some actually measured them).
Or too many people are slurping up anecdotes from the same watering hole that confirms their opinions. Outside of academic papers, I don't think I've ever seen an example of "measuring" output that couldn't also be explained by stochastic variability.
That is very, very interesting. I've been hoping to have an assistant in the workshop (hands-free!) that I could talk to and have it help me with simple tasks: timers, calculating, digging up notes, etc. — basically, what the phone assistants were supposed to be, but aren't.
"You will have to unlock your iphone first" is kind of a deal-breaker when you are in the middle of mixing polyurethane resin and have gloves and a mask on.
More and more I find that we have the technology, but the supposedly "tech" companies are the gatekeepers, preventing us from using the technological advances and holding us back years behind the state of the art.
I'll be trying this out on my Macbook, looks very promising!
The computing power we all have in our pockets is staggering. It could be tool that truly makes our lives easier, but instead it's mostly a device that is frustrating to use. Companies have decided to make it simply another conduit for advertising. It's a tool for them to sell us more stuff. Basic usability be damned.
Siri does have a setting that'll activate it if you say "hey siri" while the phone is locked. Obvious privacy and battery usage concerns though, and it's still Siri, so it's a little clunky.
I've been replacing my Google Homes and Chromecasts with Snapcast streamers, and this is the next thing I've been planning to look into.
It's truly absurd how the Google voice assistant USED to work properly for setting timers, playing music, etc, and then they had to break it 15 times and finally replace it with much slower AI that only kinda does what you want. I'm done.
Selfhosted is the way to go if you want to keep your sanity. My wife has basically given up on any Google/Apple voice assistants being able to do anything useful above "set a 10 minute timer".
> It's also possible to make an MLX version of it, which runs a little faster on Macs
FWIW, I found MLX variants to perform consistently worse (in terms of expected output, not speed) than GGUF in my measurements on my benchmark that matters to me (spam filtering). I used MLX models in LM Studio. GGUF was always slightly better.
Perhaps someone who knows more can pitch in and explain this.
It isn't 100% clear, but what quantization were you using for each? I've had worse results with MLX 8bit than what you get with Q4 GGUF, same model, seems mxfp8 or bf16 is needed when ran with MLX to get something worthwhile out of them, but I've done very little testing, could have been something specific with the model I was testing at the time.
No, they can't, not unless we get rid of the fossil fuel lobby, which pretty much runs the world these days. Which isn't surprising, given that fossil fuels are the largest industry ever created by mankind. If you compare it to anything else which was actively harmful and yet big money tried to convince you it wasn't (like tobacco, alcohol, or really anything else), there is nothing that huge. So it isn't surprising that the industry fights change.
EV adoption has been successfully held back mostly by PR, Germany shifted from nuclear to coal and gas, the US president is doing everything to dismantle anything that isn't fossil fuel and promotes fossil fuels, the list goes on.
The fossil fuel lobby can only do so much. Solar has gotten so cheap it's taking over on its own. Companies are doing it for no reason other than the math makes sense. EV batteries are nearing that point too. You can only keep BYD out of the US for so long.
Sure, but you're attributing this, deliberately or not, to the wrong cause. It wasn't that the fossil fuel industry somehow won - it was range of factors possibly including geopolitics, some existing plants aging, an emotional response to the Fukushima nuclear disaster, and the Green lobby.
Basically, they voted to kill nuclear without a solid plan for an alternative, and coal/gas is the default option for filling the gaps left in the absence of timely and sufficiently rapid investment in other technologies.
Hmm. After former chancellor (Schroeder) heavily pushed Russian gas pipelines (Nord Stream 1 and 2) and then swiftly moved to working for Russian state-owned energy companies, including Nord Stream AG, Rosneft, and Gazprom, I have a different outlook on things.
One can never discount lobbying and influence behind the scenes, but Schroeder finished being Chancellor in 2005, which was six years before the initial post-Fukushima vote in question, and even longer since various aspects of the plan continued to be supported by various politicians.
He'd be a spectacularly successful lobbyist if your suspicion is correct.
I never understood the logic behind the thinking there. Why would you ever want to place menubar items UNDER the notch, if you know it's there and they won't be visible?
It's such an easy problem to fix, with such incredible usability consequences, I just don't get the thinking.
The notch itself is probably considered temporary internally. If you code a rule for the notch, then you're going to have to consider which hardware OSX is running on in order to determine if the notch is present or not for your "notch width calculation."
PoE is not obvious to implement (take it from someone who has done it with a fair share of mistakes), uses more expensive components that normal ethernet, takes up more space on the board, makes passing emissions certification more complex, and is more prone to mistakes that ruin boards in the future, causing support/warranty issues. In other words, a bag of worms: not impossible to handle, but something you would rather avoid if possible.
I wouldn't call it "better", but the least-effort path among hobbyists and low end gear is often 12v or 24v sent over a pair with Gnd and a forgiving voltage regulator on the other end.
Really looking forward to testing and benchmarking this on my spam filtering benchmark. gemma-3-27b was a really strong model, surpassed later by gpt-oss:20b (which was also much faster). qwen models always had more variance.
If you wouldn't mind chatting about your usage, my email is in my profile, and I'd love to share experiences with other HNers using self-hosted models.
In my experience the contents of the message are all but totally irrelevant to the classification, and it is the behavior of the mailing peer that gives all the relevant features.
I hope there will be a way to plug in other models: I currently work mostly with Whisper Large. Parakeet is slightly worse for non-English languages. But there are better recent developments.
reply