Hacker Newsnew | past | comments | ask | show | jobs | submit | barumrho's commentslogin

Siri doesn't need to have conversations with you. ChatGPT can do that. But, it should be able to do actions you'd do on your phone.

Speech to text should work. I regularly have to manually edit the transcribed input. The more special words the more frequent. Completely disregards the context of the current input, for example, on Hacker news might involve special technical and IT vocabulary.

> Completely disregards the context of the current input, for example, on Hacker news might involve special technical and IT vocabulary.

Does any voice assistant do this right now? Genuine question, I don't actually know. It sounds useful as long as it's not invasive.


Any of the LLM-based ones should pull this* off - so that's to say.. none of the popular commercially available ones, yet?

Alexa+ does, but I don't use it for anything except kitchen timers and home automation triggers, so I can't speak to how well it works in a longer conversation.

Zoom's meeting notes excels at this, Google Meet is terrible at it. Meet mishears our company name about 90% of the time; various attendee names are a coin toss.

* "this" being: context consideration in speech-to-text/transcription.


Pretty straight forward on Android at least to wire up a harness that talks to Tasker[0] or another full automation app.

[0] https://tasker.joaoapps.com/


The iOS equivalent would be Shortcuts, which, while not as powerful as Tasker depending on the context, is an official Apple feature that most apps support. Claude and ChatGPT both have various Shortcuts hooks, including voice conversation.

The experience of having to tell Siri to "Ask ChatGPT <about something>" really sucks, though. It doesn't consistently do it, the handoff frequently just stalls out and you never get a response, the transcription that gets passed to ChatGPT is low quality, etc.

And though I have the feature enabled that should cause it to ask ChatGPT about things it can't answer, that works even less frequently.

But even if all of these things were true, the stuff on your phone you would expect to be exposed to the model as available tool calls, are not. So their efficacy is very limited.

(edit: iPhone 16 Pro Max, if anyone is curious)


Oh I was just thinking creating a shortcut that you'd tap on your Home Screen/control shade (whatever it's called) to activate ChatGPT, or wire up to the action button. I forgot you can have Siri do the "ask ChatGPT xyz" thing – I agree, that integration sucks.

I'd definitely do the former. I don't even think this is specific to ChatGPT or Claude's apps.

There seems to be something about how intents get triggered by Shortcuts on iOS that feels flaky to me. Whenever some app suggests a shortcut (most recently Starbucks promoted a shortcut that orders your "usual"), the success rate when I tap it is <50%.

It's possible it's uniquely worse on my device, since I haven't done a "clean install" (vs letting the device upgrade flow copy over) in like a decade. But I'm also not up for dealing with the pain of setting up from scratch just to find out it's bad on a fresh profile, either.


Is there a reason why this has to be done at training time? Could the system prompt tell the model to convert the output to a different format?


Do the ETF managers have no discretion in determining when to buy? I was under the impression that they usually handle these changes to indices gradually even under normal circumstances.


The operators of the fund are allowed to do whatever they outlined in the prospectus to track the index, some funds allow futures, options, and swaps along with equity shares to maintain parity with the index.

There are ways to gain exposure to a single stock without directly purchasing shares, options and swaps being the most common. Owning the actual shares makes things easy for the fund operators, but there are other ways.


They do, but one performance metric for these ETFs is tracking error. So they want to try to match the index closely.


of course they do. read any prospectus for a FUND and funds track INDEXES using rules. inclusion in some index doesn't hamstring anyone.

blind purchases are not going to happen. people assume passive indexing is brainless, but it isnt.


It would be cool if iPhone could double as a laptop by just adding a monitor and keyboard/mouse and switch over to macOS.


Tim Apple wants you to pay thrice, once for a phone, once again for a tablet and finally one last time for a laptop.


That’s basically the Neo.

Apple did patent a design for a dock in a monitor for a portable device to slot into. It’s gotta be getting close to expiration now. I think the trick is heat dissipation.

My friend who is a macOS programmer years ago had an idea for a startup mode for iMacs where instead of just being a screen, the storage and video card would also be accessible over the thunderbolt bus, so you could plug a laptop in and have multiple video cards at your disposal.


Thermals and the interconnection speed would be a drag, but it would be nice to have a target display mode on a mac for iPhone


100 tok/s sounds pretty good. What do you get with 70B? With 128GB, you need quantization to fit 70B model, right?

Wondering if local LLM (for coding) is a realistic option, otherwise I wouldn't have to max out the RAM.


I run gpt-oss 120b model on ollama (the model is about 65 GB on disk) with 128k context size (the model is super optimized and only uses 4.8 GB of additional RAM for KV cache at this context size) on M4 Max 128 GB RAM Mac Studio and I get 65 tokens/s.


Have you tried the dense(27B,9B) Qwen3.5 models? Or any diffusion models (Flux Klein, Zimage)? I'm trying to gauge how much of a perf boost I'd get upgrading from an m3 pro.

For reference:

  | model                          |       size |     params | backend    | threads |            test |                  t/s |
  | ------------------------------ | ---------: | ---------: | ---------- | ------: | --------------: | -------------------: |
  | qwen35 ?B Q5_K - Medium        |   6.12 GiB |     8.95 B | MTL,BLAS   |       6 |           pp512 |        288.90 ± 0.67 |
  | qwen35 ?B Q5_K - Medium        |   6.12 GiB |     8.95 B | MTL,BLAS   |       6 |           tg128 |         16.58 ± 0.05 |

  | model                          |       size |     params | backend    | threads |            test |                  t/s |
  | ------------------------------ | ---------: | ---------: | ---------- | ------: | --------------: | -------------------: |
  | gpt-oss 20B MXFP4 MoE          |  11.27 GiB |    20.91 B | MTL,BLAS   |       6 |           pp512 |        615.94 ± 2.23 |
  | gpt-oss 20B MXFP4 MoE          |  11.27 GiB |    20.91 B | MTL,BLAS   |       6 |           tg128 |         42.85 ± 0.61 |

  Klein 4B completes a 1024px generation in 72seconds.


Anyone have suggestions for a good alternative?

I've been using 1Password (family version to share some subset within the family) for more than 10 years now, but I have to say the user experience has degraded quite a bit. Anyone have a better overall alternative? (Doesn't necessarily have to be cheaper.)


Using Enpass, migrated from 1Password when in need of a Linux client some 10years ago. As early user I was grandfathered into a free lifetime account and eventually was required to pay a discounted lifetime fee $70 through Apple, which I’m fine with, it’s Indian developers need to pay their rent too.

Enpass has all features I need, on all platforms including iOS. It syncs using the api of one of the free storage providers, WebDAV or even over WiFi. Having some 600 entries and a few attachments (copy of ID Cards etc) and never had any performance issues. Nor issues with subdomains. Regular updates, most recently added PRF (Pseudo-Random Function) for passkeys. It lacks a command line client, which I can live with. Nor does it support the fingerprint reader on Linux, instead has a pin option for quick unlocking.


I've used 1PWD for at least as long as you, and when renewal comes around (EDIT: oops, guess I never "upgraded" to subscription plan) I'm going to cancel and just stick with Apple's Passwords app (née Keychain Access). First "cloud!", subscriptions, now 33% price increases for the hell of it, I'm outta here, 1PWD. (Though in looking just now, we never upgraded to v8.0, so I guess I'm already outta here.)

I only suggest Passwords because if you've used 1PWD for that long, odds are good you're on Apple HW/OS. It does everything we need in our household, including shared creds. One of these days I'll get off me arse and export the 1PWD stuff (IIRC, 1PWD->Apple PWDs is doable). Right now we use 1PWD as R/O, and all new stuff goes in Passwords.


Good to hear that is a viable option. I am mostly on Apple devices, but do need Linux and Windows support, so I don't think it will work for me.


I'm working on an alternative that I hope would be better. https://github.com/lockstepvault-hq/lockstep (early alfa project)

Would you mind sharing what user experiences are not ideal with 1Password, I'd like to know I can address those those in Lockstep.


I'd say it's mainly to do with browser/iOS plugins not being responsive. I find myself often resorting to opening the app and copying and pasting the password or other info because autofill function doesn't work on different websites.

Otherwise minor UI things like categories on the sidebar which made it easy to navigate, but they got rid of it a while back.

Good luck with your project!


If you're in the Apple ecosystem, password and passkey management via iCloud is included.


Don’t lock yourself in apple ecosystem especially in such important thing as password management.


You can export from their app very easily.


Can you export it once apple locks your account?


I doubt it. But I keep my own encrypted backup anyway (as I did with 1P, too), so realistically only the most recently added/updated passwords are at risk.


Why not? How is the risk substantially different than with a different vendor?


First, all your passwords are locked behind the same passcode you use to log-in. You cannot set different password to unlock the passwords. Anyone who saw your iphone passcode or macbook’s password can not only onlock the device but get access to all the passwords as they are behind the same passcode.

Second, what would happen if you were locked out of apple id account? Or don’t have access to apple hardware.

Password manager should really be platform and device agnostic.

That’s why people used stuff like 1password in the first place.

You really don’t want to put all your life into a single account (that’s why you should not use sign with google or what have you).


People (at least I) started using 1pwd because it was substantially and amazingly better than keychain at the time.

The delta has greatly decreased.


Bitwarden, keepass


Bitwarden. You can host a free instance with Vaultwarden.


While I agree that the return numbers are probably very low, you may be underestimating Reddit's impact in terms of product recommendation. I noticed that Reddit results are often pretty prominent on Google search when product reviews are searched for. Security camera market is pretty competitive and a single factor like this could easily sway people to choose alternatives.


Apple Music isn't native either.


Oh right, I forgot, "native" just means "good". So if an app is bad, it can't be native, and if an electron app is actually good its because they're doing crazy optimizations that aren't feasible for mortal souls so don't even think about it. This is the "Hackers News Law of Application Nativeness".


AI coding tools are providing non-programmers giving similar ability. Not production quality, but still useful to them everyday and they can tweak as they go.


“이/가” is the Korean version of this. It depends on the sound of the word that precedes.

I still remember seeing it when I first started using Windows 95. As a kid, I was amused that it didn’t know which one to use. Really, I didn’t even know that I was making that choice (and couldn’t say what the rule was).

If anyone is interested about it, this page explains https://www.90daykorean.com/i-ga-grammar/


Haha, interesting!


Consider applying for YC's Summer 2026 batch! Applications are open till May 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: