Rewind.ai now available for iPhone | Hacker News Reader

Rewind.ai now available for iPhone

apps.apple.com

92 pointscdolan65 comments

It’s been 14 years since my first submission to HN and I am so grateful to this community for all your feedback & help on my first company (Optimizely) and now my second one (Rewind). Today we’re introducing Rewind for iPhone - a truly personalized AI in your pocket. I’d love your feedback!

Browse & search for any word you’ve seen (including screenshots)

- Rewind automatically captures what you read in Safari and imports your screenshots.

- You can preserve & search for anything you’ve seen on your iPhone.

- This opens a new dimension of computing: time.

Summarize and ask any question using AI

- Rewind integrates with OpenAI GPT-4 to allow you to ask any question about anything you’ve seen.

- Create summaries, synthesize information across apps, or remind yourself how you know someone.

Private by design

- Private browsing in Safari is not captured.

- Using GPT-4 is optional and only the relevant text necessary to answer your question is used.

- Your data is not used to train AI models.

- All recordings are stored locally and only you have access to them.

Learn more: https://www.rewind.ai

Please let me know what you think!

What data leaves my iPhone?

I have been a Rewind user on the Mac for months and on the iPhone (beta) for a few days now, and there are basically no tools in the past few years that have changed the way I work this much.

I get a transcription and summary for every meeting that I'm in. The summary fairly reliably captures all main points, and definitely enough to job my memory on anything else. It gets action items.

I use Ask Rewind constantly – "who mentioned X in a meeting last week", "what did Joe say was his top priority for Q3", "was I reading about project X in Notion, Slack, or Jira last month" ... all questions Rewind handles well.

It just makes me a better colleague. I am a bad note taker, and this allows me to be more present and engaged in meetings because I have zero stress about forgetting something.

It's not perfect by any means; I wish transcription was more accurate and things like that, but all it needs is marginal improvement in a bunch of areas and we're so early. Such an exciting product.

> I get a transcription and summary for every meeting that I'm in.

So it records and transcribes audio and stores it on your phone?

That's awesome! But I wonder what my corp security team would think about that. I would however love to have such a thing because I too am a bad notetaker, as I prefer to listen closely to what others are saying.

I believe the meeting transcription is Mac only currently and can be set to automatic (via syncing your calendar and it detecting the meeting automatically) or manual (toggling on and off mic recording).

I do think this type of thing probably drives compliance and security folks mad and will be a barrier to adoption in many companies. All of my work is client-based rather than traditional FTE which largely gets me out of that sort of thing.

Not to mention two party consent laws about recording conversations.

https://recordinglaw.com/party-two-party-consent-states/

I am responsible for crafting and implementing GRC and infosec policy at a fintech. We review and approve AI tools with DLP, auditing, and logging wrapped around them. Meetings require a click through to acknowledge when they are being recorded.

I used Rewind for a few days but stopped after my resource usage spiked. I didnt research much into this; but have you seen any perf issues when using Rewind.

It has gotten better, but it's still a factor. Relatively shortly after launch they switched all processing to only occur when plugged in vs. on battery, which helped from a consumption standpoint.

As far as pure performance I can't speak to it too much; I have a massively overpowered machine (M1 Max w/ 64 gb RAM) for my typical workload (meetings/email/docs) so I don't feel anything.

The screenshot feature is achingly close to my dream bookmarking app. - screenshot something you want to come back to (a tweet a song on spotify a youtube vid) - "AI" figures out what it is and gets the relevant info (song/vid title, URL, which app you're using, etc.) - files the item appropriately in a nice searchable way - unknown things get put in an inbox for you to come back to and file yourself

imo until devices get more bold about how they manage your experience (ie watching everything you do with on-device AI to be your full-time assistant) then screenshots are the best vector for this kind of thing.

anyway, exciting to see folks trying to get into this!

Part of me wants to start 24/7 recording of my desktop now for later use when this kind of tech matures. I imagine eventually can get away with recording at fraction of resolution and FPS, color info, sound bitrate etc and have AI upscale/sample everything. Imagine recreating documents by prompting to turn segment of video scrolling through content and have AI split out a pdf.

So it's a reader app for what happens on you mac plus a Safari mobile browser extension to record what happens there? Clever implementation for an OS that is limited in what it allows in this regard.

I've tried Rewind on the Mac for a little bit but turned it off because I thought it was causing an issue with my notifications (I think now it's related to something else). I never "used" a rewind in part because I hate using consumable things (See also my slight aversion to Kagi, though I want to give it another shot). I also dislike pricing that's not really based on ongoing costs (searching your local archive 10 times or 1000 doesn't seem to cost Rewind anything). I was also not super excited about the whole "we will summarize your meetings" feature I got a notification about as it requires sending your data to Rewind. Also, I think part of my issue is that while I like having the history I don't have the muscle memory to use said history yet and so I forget it's there.

Thanks for the feedback!

On Mac: you are right that recording audio causes your notifications to be hidden by default. There is an OS setting to change that behavior. Here's how to change it: https://help.rewind.ai/en/articles/7039599-what-are-the-limi...

On pricing: you are also right. We are thinking about new pricing & packaging that feels fairer. We're thinking about doing purely feature-based package differentiation (as opposed to tracking number of "rewinds"). Would love your thoughts or suggestions.

On muscle memory: again, you are right. Some of our users think of our product today as mostly insurance and don't actively use it every day. We hope to change that by giving them valuable things that fit into their existing daily workflow (e.g. summarizing meetings, telling you how you know the person you are about to meet with, proactively drafting emails for you based on context)

I don't think you can access your recordings from your mac on your phone as far as I can tell.

Totally agree on the muscle memory part. I’m hoping this can turn into the second nature that searching google has become.

Very excited to see them get this out there. Rewind on Mac is useful, adding iOS would be all the screens for many people.

That being said, this is not the same as the desktop version. It does not record you're screen constantly. As far as I can tell it is capable of recording Safari, and it can source screenshots from the Photos app.

Not complaining—Safari support is probably the 80% of what I'd want to recall from mobile.

And really it's Apple you need to complain to if you were complaining. I'm sure Rewind would work exactly the same on iOS if Apple allowed it.

Indeed. It's horrible that this is not allowed even if the user wants it.

“Private by design” is a killer feature for me. This kind of app could be absolutely nightmarish if all the data were stored and analyzed remotely. Thank you for building it this way!

It's powered by GPT-4 so it's not "private by design" by standards here:

https://help.rewind.ai/en/articles/7791703-ask-rewind-s-priv...

From the post I replied to:

"Using GPT-4 is optional"

Yeah, if you opt in to using a cloud service, you'll be using a cloud service.

Optional for what? If you use Ask Rewind you currently need GPT-4, and AI is the meat of this product.

I think they shouldn't tout privacy as a feature until their models go open source. If they did that what would they say then, "Now we're really private"?

Can this retroactively retrieve screenshots I've taken of Safari over the past? I've made a habit of screenshotting interesting bits of articles and discussions online, and now have the photographic equivalent of a gigantic unsorted bookmarks folder, just also mixed in with the rest of my camera roll.

The Photos app already automatically OCR everything and you can search through any images including screenshots.

True, but like my gigantic unsorted bookmarks folder, I'm not really sure what I want to find.

I like the idea of an AI tool that could help me not so much search specifically within my screenshots, but rediscover what I was recording in the first place.

Yes, just tried this!

Yup

One of the key things for many people will be: Are you vacuuming up all my data?

I would imagine this probably vectorises a lot of data and stores it in a vector db, I'd be surprised if that's all happening on my hardware, happy to be wrong though.

I've been working on a similar system for a while called irchiver [https://irchiver.com/] but what I wanted from the start is for everything to be local and plain formatted, i.e. stored as .txt and .webp to be most accessible.

Coincidentally, I've only been building for Windows, so different from the platforms supported by Rewind. If anyone wants to collaborate, I'd be very open to it.

would love to hear more about this!

So I was thinking about how the web is not as permanent as it used to be. Between walled gardens, and dynamic content, it's hard to find things you saw. So I wanted a way to have a screenshot-based "archive.org" but for yourself, so that it works with non-public content too like a Facebook post or if it disappears like all the stuff on Google+.

Philosophically, I think almost everything we see and think about is in a web browser these days. Even private conversations with my friends, I often look up a few things about them. So I've been using irchiver myself to dogfood it, for almost 2 years now. And found it to be super useful already. For example, I've recovered important technical posts that someone deleted, paragraphs on text I lost when submitting a form that didn't save, or found a comic strip that I had at the tip of my tongue but got lost in google search.

And I think it's now possible because of strong+efficient screenshot compression and good OCR. There's an undocumented Win32 hack that lets you capture window content in a fast way (moreso that a typical BitBlt) so I used that to grab the images. And made a standard inverted index search engine for the content. Anyways, the webpage I put up for the project explains it as best as I can to a general audience [https://irchiver.com/].

In many ways, it's like the Rewind startup here. Except it just happens I did it for Windows, and they did it for MacOS and iPhone. And I care a lot about local storage of the original screenshots and text. Would love to hear others' thoughts.

Can you say more about what “using GPT-4 is optional” means? What features do I lose by not using GPT-4? How do I opt out?

You would lose all of the natural language q&a (ask X) which imho is the useful piece.

Do they send any data to OpenAI or any other cloud based LLM?

At least on OS X, yes. All of the recording/storage is local, but when you ask any kind of questions involving AI, it sends relevant pieces of history (as text) to OpenAI. This may have changed, but it was the case when they launched the beta.

They say absolutely not, and that everything is processed and stored locally.

*edit* came back to correct this and see that I have already been corrected. If you ask Rewind a question it explicitly says your question and ‘relevant’ web snapshot are sent to ChatGPT 4

Sorry for jumping the gun

I'm a Rewind user and really enjoy the product, but this isn't quite true. Again, maybe this has changed since the beta, but the "Ask Rewind" page has the relevant info[1]:

> No audio, image, or video files ever leave your machine. Ask Rewind only sends relevant text-based data to the cloud, excluding personal sensitive information such as credit card and Social Security numbers.

> When you ask a question, only the text of the most relevant moments from your past are sent to GPT-4 to generate an answer.

[1]https://www.rewind.ai/ask-rewind

Usually what is “relevant” is determined by vector search. So user has no control of what is being sent out. Is that how rewind work?

That’s what make me worried.

You can choose not to use the GPT-based feature, that is a recent addition, not really the core product

They say absolutely yes:

https://help.rewind.ai/en/articles/7791703-ask-rewind-s-priv...

This is such a cool application of AI. If it's sending everythign to some cloud provider though it's a privacy nightmare. Would love to see in a few years the option to self host models like this on a VPS or even your own hardware directly.

They already do everything locally, no wait needed.

Just for storage lol, all the AI is cloud still.

Not any of the AI features (ask/query/etc.).

I've been using Rewind for a while and got a chance to try out the iPhone implementation when it was in development. I have to say, the Rewind team is pretty creative in how they thought about moving the same hands-off experience from Mac to iPhone. Personally most of the stuff I want to Rewind to is in Safari anyways so this is perfect to me (and the screenshot import fills in the rest, though I have a ton of junk screenshots of my lockscreen...).

I hope one day we're able to combine our Mac & iPhone indexes (and privately!), since that's the last cognitive load I have (which computer did I look at something on, or was it my phone?) when using Rewind

Ha I totally have the same problem of remembering which device I looked at something on!

> Private by design

How are people supposed to believe you if your product is not open-source?

Wireshark? Little snitch?

(I work at Rewind) We even have a page talking about using Little Snitch for our mac app https://help.rewind.ai/en/articles/6831953-why-does-rewind-m... We'll have to update for our iOS app

Is a behind-the-scenes engineering blog post of the trade offs you made for iOS when compared to the macOS app available?

Have been following your application and I think it is a wonderful application of AI and this new exciting LLM thing. Sadly I'm not on any of the platforms that you have currently supported.

It would be very cool if someone can create an open-sourced version of this for Linux users :)

Is there a limit to the amount of activity you can do on the iOS app? Everything was working great for the first few hours that I had the app but it has stopped tracking anything.

This looks like a fantastically useful product. In my mind this is the type of product that suits platform builders. How are you thinking about moat-building.

The safari part sounds interesting. Searching screenshots for text is something you can do from the Photos app already.

https://support.apple.com/guide/iphone/search-for-photos-iph...

I love this app! The combo of Screenshot/Safari browsing and AI Chat is incredible... It makes screenshots contextual and useful again!

Don’t know why I thought this was going to be an app to let me downgrade iOS on my device. Clearly I need to step away from the screen.

I want something like this but for the whole display. Due to the security implications, obviously it would need the OS native support.

Just installed it. Seems like it requires you to browse using Safari (because of extensions)?

The good news is technically you're probably already using safari, even if you don't realise it.

Nope. The test page fails for me.

At least in MS Edge: https://safari-test.rewind.ai/

It's only going to work in the main Safari since that's the only thing that supports extensions.

WebKit yes. Safari, no

rewind.ai feels like a super power.

so. HN folks are famed for privacy consciousness. what should we be looking out for here?

Audio/video is stored and transcribed locally (https://help.rewind.ai/en/articles/6526621-who-has-access-to...) but they send all your text (including the transcribed audio) to OpenAI (https://help.rewind.ai/en/articles/7791703-ask-rewind-s-priv...)

There is no privacy until they eliminate the GPT-4 dependency.

Reply on news.ycombinator.com