Show HN: EchoStream – A Local AI Agent That Lives on Your iPhone | Hacker News Reader

Show HN: EchoStream – A Local AI Agent That Lives on Your iPhone

22 pointsshuhongwu27 comments

Hi HN,

I’ve been working on a project called EchoStream, and I’d love to share it with you.

As an AI startup founder, the first thing I do every morning is open Hacker News. I look for what other founders are building, check out new GitHub projects, and explore new products. But I started noticing that it was taking me the whole morning just to get through it all.

I began to wonder: could an AI help me read Hacker News more efficiently? Could it summarize everything so I can quickly scan through, and then dive deeper into what interests me? And more than that—could it store what I’ve seen so I can recall it later when it's relevant?

That’s the idea behind EchoStream.

EchoStream is a local AI agent that runs entirely on your iPhone. It includes a built-in thinking model to help you do more with less effort—privately and efficiently, without sending your data anywhere.

Some of its key features:

- Read and understand web pages using AI

- OCR for recognizing and processing text in images

- Automatically record, transcribe, and summarize audio

- An AI-powered news feed with fresh, relevant updates

- Memory search that connects new info with what you’ve seen before

It’s still early, but I’d really appreciate your feedback or thoughts. I built this for myself, and now I want to share it with all of you.

You can have a try here:https://apps.apple.com/us/app/echostream-mind-copilot/id6476... or to the homesite: https://www.echostream-ai.com/

Note: It works well on any iDevice with at least 4GB of RAM.

While this specific app may be paid, I built https://HackYourNews.com (free) specifically to "summarize HN for me" and it grew to include many other websites. Uses Gemini Flash under the hood and summarizes both the story AND the HN comments!

That's really cool! I just checked out HackYourNews.com and it's a great implementation of the summarization concept. I love that you're covering multiple sites beyond just HN, and including the comments summary is smart - often the real insights are in the discussion threads.

It's interesting that we both saw a similar need around staying informed without spending hours browsing. Your approach with Gemini Flash for the web-based solution makes a lot of sense, especially being able to offer it for free.

Our focus is a bit different - we're exploring what becomes possible when you move that intelligence directly onto the device itself. So while you're solving the summarization problem really well in the cloud, we're asking "what if this kind of AI agent lived natively on your iPhone?" The privacy, speed, and offline capabilities open up some interesting possibilities.

Thanks for sharing your project - it's always great to see different approaches to similar problems. The fact that multiple people are building in this space suggests there's real demand for better ways to consume information.

And, We will remove the paid feature in the later versions. Cause this agent is running on your device, there is no need to have any pay.

Very cool. Some ideas;

- would love to collapse all the news outlets at top and keep closed - would love a way to customize the wrap ups below that you have - would love the comments summary below

Great ideas! We're totally considering these.

For the comments summary - honestly, we think comments are pure gold. Sometimes they're way better than the actual news (which says something about journalism these days ). I personally read every single comment because you never know when someone drops absolute wisdom in there. But, not everyone has my comment-reading obsession, so we'll definitely add a summary option for the normal people!

This is cool, I always like to see local-first services on Mobile.

All of these features are great, and it's really cool that I can have them run locally on my phone, but what I feel is really missing on the Mobile App market right now is AI cross app automation.

I'm not sure it's possible with IOS right now though.

For instance: I went to france earlier this year, and before I went I saw this french vocabulary sheet that was just an image. I wanted to get the text from the image, format it into a csv with english in column 1 and french in column 2 and then export it to anki, which is also on my phone. AnkiDriod has this feature, and Gemini can deffinitely do OCR, translation and text formatting, however I had to do all the awkward app switching and input myself! This is the bit I wanted automated

In my knowledge domain, that's actually possible on iOS, but it will cause more privacy issues for sharing data between different apps, and also the clipboard issue. Apple has been pretty restrictive about inter-app communication for good reasons, but it does limit these kinds of automated workflows.

As you mentioned with that situation, I think Apple and Google may have better power to optimize the OS system if we only do it on the local device. They have the deep system-level access that would make this kind of seamless automation work without the privacy and security trade-offs that third-party apps face.

What about an agent that picks the best idea, steals it, runs the business for a while, goes bankrupt, and then emails you an executive summary of bitesize "learnings" from the exercise.

You can't summarize your way to success! If you're spending more time clicking and reading filler words than you are thinking, then you're doomed from the start.

Our brains summarise all the time.

I can't remember exactly what texture/shade is on the ceiling of my bathroom, but I remember roughly what colour it is. If you asked me if there's a bit of pain flaking off and where it is, I couldn't tell you to save my life.

When I open HN (or most things) I will skim down the list of titles and it's keywords that will jump out at me.

When I read threads I will start skimming when the conversation evolves into endless debate, unless I'm interested in said debate.

Processing every possible bit of input in full would not make for a very efficient or worthy experience, but of course this is contextual. Someone a contract should read every word; though this has been weaponised in the form of long ToS - they know most people don't read those at all.

Do you have any insight / resources into developing a local AI?

I have a novel use case as a wilderness first responder and am looking for a pathway.

1. llama.cpp can run on both Android and iOS devices. 2. For non-transformer models, Mamba and RWKV are also good options. 3. Additionally, you should have a better understanding of embeddings, tokens, and the structure of transformers.

And on Apple Silicon chips, MLX is preferred and is often the best choice.

Thanks!

I’m trying to create something that will run online when connected, run local when offline, and dynamically download relevant files to prepare for offline use.

Not available in my region, apparently.

We're currently only available on the App Store in the US while we're in early development.

You could try the TestFlight version if you're interested in testing it out - that should work in more regions.

Where is the TestFlight link?

Here:https://testflight.apple.com/join/x18p4hz7

Better have an iPhone with RAM > 4GB

This is the feature that I am interested in

- Memory search that connects new info with what you’ve seen before

Yes, this is like a kind of flow - how to say - a kind of mind stream, replacing me thinking of any connections between any newest news and my memories.

As a person, we only need to listen to some insights and decide what to do. Others, let the ai agent do it. We think this may be the future in personal AI area.

Actually, imagine your device(not only the iPhone but also other edge devices) constantly building these connections in the background - when you read something new, it automatically surfaces related things you've seen before, conversations you've had, or notes you've taken. Instead of you having to remember "wait, didn't I read something about this topic last month?", the agent just presents those connections.

It's like having a research assistant that never forgets and can instantly connect dots across everything you've ever consumed or created on your device. The human stays focused on the insights and decisions, while the AI handles the memory and pattern recognition.

Seems to be region locked? "Not available in your country" in Australia.

Just set up for opening to Australia, it may take some time to wait.

I’m in the US but it’s saying not available in my region?

Is it possible to download the app based on your App Store account's region, may be your account is registered outside the U.S.?

Also the app's age rating is 17+.

Is this not available in India? Would love to try it.

Yes, we’ve just made our service available in India.

Why make it not available in all regions?

We're still developing this product for tech leaders and startup founders. We're also working on AI products for edge devices. Expanding to other regions would create too much pressure.

If you like this idea, I made something similar at https://saguaro.thelinell.com, although it works using various RSS feeds.

Reply on news.ycombinator.com