What value to you get from having this happen on every turn with a change? I get that the ratio should be a lot higher than the default. Doesn't doing this for every turn where something changed slow things down a lot without gain? I haven't experimented, that's why I'm curious. Why not 1 line or 2 in a CLAUDE.md or AGENTS.md so that the agent ups the frequency, but doesn't make it every single time?
1 line in CLAUDE.md I tried (one line, 100 lines, many iterations) - it not always follows style instructions.
The value is very large for me - instead of reading page of Claude produced prose, I mostly just look at focused choices and continue with single button press - significantly less Claude deciphering and typing now. In addition, I do my sessions with /rc so I get nice form on mobile app where I can just select a button instead of typing to continue work.
PS: I think there is confusion in term - "every turn" - I don't mean every tool call, I mean after minutes or hours of agent work when agent decides that its done it gives a report. The report I see now has structure to it with options set to continue (or stop).
Congrats on the launch! I'm glad more progress is being made in this area.
Because of the hallucinations inherent in transformer models, I went looking for a transducer-based model with a low WER. I have been super pleased with parakeet-unified-en-0.6b. It's WER isn't as low as Canto, but it's about as low as you can get (~5-6.5%) with a non-transformer-based model as far as I'm aware.
I've been very pleased with its output.
I wasn't looking for this, but it's also lightweight enough to run on my little potato PC, which has an i5 8th gen processor, and still transcribe 9-10x faster than realtime.
I vibe-coded a little wrapper for it and use it on folders of audio or podcast rss feeds or even YT playlists and channels and it's been one of my new favorite tools.
I used Snapdrop before they got bought and found it pretty reliable. It sometimes would not transfer and I went exploring.
I used magic-wormhole-based Warp and liked it, but now I use croc of getcroc.com for almost everything. Whether it's sending files to myself or others, it's just the most flexible and seems to be reliable and fast from my use.
(I haven't been using it super long, but I heartily recommend.)
I like croc but you can get magic wormhole clients for platforms that your SO uses through the usual distribution channels without having to resort to "weird" installation methods like Obtanium or Fdroid.
I've never used Iroh; can you use it with no setup or without an account?
I send/receive files often enough from non-technical people, having a no-install/accountless solution for sending files is essential. For my use-case, my go to is croc or Bitwarden Send (I'm on a paid acct), depending on the person.
Update: I see that there are many ways to do this accountless. For some reason the search result was iroh.computer and I made some assumptions. There are definitely some iroh-based options that look like they would have essentially the same UX that I'm going for. Would love to hear if there's some specific pros or cons to croc or iroh.
"These new capabilities are powered by the next generation of Apple Foundation Models, custom-built in collaboration with Google and its Gemini models for deeply integrated Apple Intelligence experiences. These latest models run on device and on servers using Private Cloud Compute."
TIL about <meta http-equiv="X-Clacks-Overhead" content="GNU Terry Pratchett"> which is delightful and an example of what I love best about the Internet.
IANAL, but it might be relevant to others trying to answer what jurisdiction you are in. (US, EU, CN, JP, elsewhere?)
Also, when you bought from PianoDisc, did you agree to abide by a certain jurisdiction's laws in your use of PianoDisc? And did you explicitly agree to not share any sort of decoder/encoder in any ToS?
> (k) to interfere with or circumvent the security features of the Service
IANAL. But I think reverse engineering their data structure, identifying a security measure - even one as weak as obfuscation, and publishing code to circumvent it is clearly against your license.
I am also NAL, but out of curiosity does OPs post detailing the obfuscation transgress any laws?
If the security measures exist in plain sight, as they apparently do, are they allowed to be discussed?
I reckon that if OP posts the encoder/decoder software that'd be against some sort of license clause. However in the age of AI who cares about the software at this point? Anyone can prompt their own private version into existence.
Just thinking out loud here. I have not considered AIs use as personal "cheat engines".
Well, it already ranks pretty highly in Google search results, so it’s unlikely that Elon’s lawyers haven’t heard of the website.
It has some page where it almost goes out of its way to say that it doesn’t bypass security measures like Nitter and XCancel supposedly did. So Elon might not have even a fig leaf of a legal argument to take it down like he did with the other websites.
It’s also possible that Twitterviewer is just flat out disobeying cease and desist letters from Musk, which nitter and xcancel weren’t willing to risk.
In any case, I find it unlikely that Musk hasn’t heard of Twitterviewer.
reply