Given how the blog is presented, I assumed this was something novel that solved a unique problem, maybe a local multi-modal assistant for your device.
I installed it and it's none of that. It is a mere wrapper around small local LLM models. And, it's not even multi-modal! Anyone could've one-shotted this in Claude in an hour (I'm not exaggerating).
What's the target audience here? Your average person doesn't care about the privacy value proposition (at least not by severely sacrificing chat model's quality). And users who do want that control can already install LMStudio/Llama.cpp (which is dead simple to setup).
The actual release product should've been what's described in "What's next" section.
> Instead of general chat, we shape Ensu to have a more specialized interface, say like a single, never-ending note you keep writing on, while the LLM offers suggestions, critiques, reminders, context, alternatives, viewpoints, quotes. A second brain, if you will.
> A more utilitarian take, say like an Android Launcher, where the LLM is an implementation detail behind an existing interaction that people are already used to.
> Your agent, running on your phone. No setup, no management, no manual backups. An LLM that grows with you, remembers you, your choices, manages your tasks, and has long-term memory and personality.
> Anyone could've one-shotted this in Claude in an hour
I think they did. If you start the download and then open the sidebar and/or background the app, the download progress bar disappears and is replaced by the download button. If you press the download button again, the progress bar reappears at the correct point.
I find that Claude often makes little statefulness mistakes like that. Human developers do too, but the slower and more iterative nature of human development makes it more likely that that would get caught.
> Anyone could've one-shotted this in Claude in an hour (I'm not exaggerating).
This probably could have been one-shotted with Sonnet, not even Opus. Given how over indexed they are on LLM coding, Haiku might even be able to do it.
This is actually an interesting coding model benchmark task now that I think about it.
I hate to say it, but this looks like the sort of thing a CEO told their team to build on Monday morning in a panic because they are grasping for ways to participate in the AI craze. And the team did just that: they built it that morning using Claude Code.
There is truly nothing original here and the product doesn't have a chance in hell of earning money. Local LLMs on-device will be dominated by the device vendors, whose control of the hardware stack combined with their ability to subsidize billions of dollars of machine learning research gives them an unfair advantage. Apple knows what the next generation of silicon will deliver, and their ML engineers are already hard at work building models that will be highly optimized for that silicon a year or two ahead of time. Open source models are really great and are backed by well funded labs; however, delivering these models on-device in a way that pleases users will never be easier than it is for the vendors of the devices.
Plus, device vendors have ways of making money from local LLMs that third-party app providers do not. They can make their local LLM free and earn money on the hardware play, without skipping a beat on the billions of dollars of ongoing R&D. I don't see how third party app vendors make money here when they will be competing with the decent, totally free alternative that Apple and Google (and Samsung etc.) will load on in the next year or two.
> We have not seen a tidal wave of untechnical people vibe coding up their own software solutions.
When my little brother who is a drummer, and has never even looked at "code" before, had claude on-shot a python app that let him download songs on youtube, extract the stems, collect tempo/key/etc information, then feed that into his DAW, all without ever looking at code, knowing what any of it did, etc., I knew that we were about to see a LOT of single-use applications.
I'm not against it, honestly. I have always written little one-off scripts and apps that accomplished something faster than manually, now that those one-shots are possible with an LLM in seconds sometimes, it makes all my personal scripts so much easier... that said, I definitely read the scripts that are output, and attempt to step through them in a debugger before assuming it is all good.
Probably just another ollama-type service who wants to slide themselves in between the user and local models, so they can take all the credit and work on convenience based platform lock-in, then later introduce paid tiers.
I would love to see a "distributed LLM" system, where people can easily setup a system to perform a "piece" of a "mega model" inference or training. Kind of like SETI@home but for an open LLM (like https://github.com/evilsocket/cake but massive )
Ideally if you "participate" in the network, you would get "credits" to use it proportionally to how much GPU power you have provided to the network. Or if you can't, then buy credits (payment would be distributed as credits to other participants).
That way we could build huge LLMs that area really open and are not owned by any network.
I would LOVE to participate in building that as well.
Maybe I’m missing it but the page is really light on technical information. Is this a quantized / distilled model of a larger LLM? Which one? How many parameters? What quantization? What T/s can I expect? What are the VRAM requirements? Etc etc
Huh, 1.6B/2B/4B models, I guess they weren't joking when they said "not as powerful as ChatGPT or Claude Code". Also unsure why they said "Claude Code", it's not an CLI agent AFAIK?
I so wanted to love Liquid AI's models, but despite their speed I was never able to get anything useful out of them. Even their larger models can't be trusted with simple stuff like inserting a column into a markdown table. The advertised tool calling is also not great. What I found interesting was that the ones I tried were a little light on guardrails.
I would really like to know what people use these small and tiny models for. If any high-karma users are reading it, would you consider posting Ask HN?
> otherwise small models can be very effective within the right use cases and orchestration
very limited amount of use cases, perhaps. As a generalized chat assistant? I'm not sure you'd be able to get anything of value out from them, but happy to be proven otherwise. I have all of those locally already, without fine-tuning, what use case could I try right now where any of those are "very effective"?
I tried it on my iPhone 13 mini. I believe the model you get changes depending on your phone specs. For me it downloaded a ~1.3GB model which can speak in complete sentences but can’t do much beyond that. Can’t blame them though—that model is tiny, and my device wasn’t designed for this.
There's dozens of local inference apps that basically wrap llama.cpp and someone else's GGUFs. The decentralized sync history part seems new? Not much else. But the advertisement copy is so insufferably annoying in how it presents this wrapper as a product.
Have a comparison chart to Ollama, LMStudio, LocalAI, Exo, Jan.AI, GPT4ALL, PocketPal, etc.
There are so many wrappers that are obviously wrappers. I wonder if part of the value proposition here is that it is “like a product.” I have no idea if they actually achieve that, though, and doubt it really could be proven on a site.
I like the idea of having a user-friendly app that lets you use LLMs locally. Tools like Ollama and LMStudio tend to put most people off because you have to decide for yourself which models to use and there are so many settings to configure. If the hardware you’re using is compatible, Ensu could be a drop-in replacement for casual ChatGPT users.
However, it’s a bit confusing because, for example, a larger LLM model was downloaded to my smartphone than to my computer. It would probably make the most sense if the app simply categorized devices into five different tiers and then, depending on which performance tier a device falls into, downloaded the appropriate model and simply informed the user of the performance tier.
Over time, it would then be possible to periodically replace the LLM for each tier with better ones, or to redefine the device performance tiers based on hardware advancements.
Heard the first time about them (ente) yesterday in a discussion about "which 2FA are u using?". Directly switched to https://ente.com/auth/ on Android and Linux Desktop and very happy with it.
You presumably had a working 2fa app already, but off the cuff decide to switch to new unvetted variant X; basically unknown auth system after reading a few paragraphs of text in an afternoon?
Ente is extremely well known in the privacy circles, so this is not just some random company with a random app out of nowhere.
Check PrivacyGuides for example.
Here’s where it was added to PrivacyGuides - https://github.com/privacyguides/privacyguides.org/issues/36.... The person opening the issue is the CEO of ente. So the CEO of ente gets his company mentioned in PrivacyGuides back when it was new and that makes it more legit?
PrivacyGuides goes through their own process of vetting (whether you would agree with their process or not that’s another topic) so I think the discussion to add Ente Photos is the more relevant link
https://discuss.privacyguides.net/t/ente-photo-management/11...
While I would have the same reaction, in this case I think it is a sane decision. Ente is cornering the privacy market and I think they're doing a great job. They have a lot to lose (trust) and it would be stupid if they did something shady with the data entered in the 2FA app.
Not knowing them, how could OP trust them instantly? Whether they really have that trust or not, you have to know them for a while and from many different trustable sources. The story is a bit strange.
They just store tokens, without other FA at "worst" you get locked of your account but nobody else has access either. You're also supposed to, as good practice, not be limited to token generation and typically have a dozen or so of recovery tokens. Also if they were somewhat not working at doing the 1 task they should do, namely generate tokens, then you won't be able to use them so it won't even be added.
So... I might be missing something, can you please explain what worries you and why I should thus worry too?
Not saying they’re a paid promoter. But if I paid someone to speak about my newly launched product, they’d say something exactly like that. “Never heard of these guys before, but I loved their other product you’ve never heard of. I’m super excited to try this one!”
I am not a bot and I am not associated with this company in any way. But I am a happy user of Ente Auth as well. This AI thing they made however just gives off "we have to do something with AI or we'll be left behind" vibes.
agreed. i have never seen anyone (let alone an assortment) of hacker news users saying "i switched my 2fa to this after seeing how great it was!" Not really sure how one 'switches their 2fa' to an LLM...
This thread is about the 2FA app, not the LLM app. I don't care about the LLM app. What's this witch hunt? This app literally solved a (self-inflicted) problem I was having for some years now where I was keeping an old phone around just for MFA. I even thought about creating an iOS app that's compatible with Aegis files (actually I even _started_ working on that, but didn't get far) just to solve my problem. Now I don't have to, thanks to a comment here, and that's why I posted. Geez. I guess I'll stay with negative comments for the future, they seem to be more trustworthy.
I mean I get it, astroturfing is a real problem and an annoying one for communities. But I also have no idea how to prove to you that I am neither a bot nor shilling here.
I'm very happy syncing between KeepassXC on Debian and Keepass2Android on mobile. It handles TOTP accoss devices.
What I'm missing is a way to create and use Passkeys across devices. My use case does not support creating a new Passkey on every device, I need to sync them via servers I control. The system that supports that will be the system that I migrate to.
I was just thinking their end goal seems to be to harvest creds by putting their own rebadged distribution of local models. That’s the only “business” model that makes sense.
Expressly harvesting creds through a 2FA app seems a little more direct.
Oh, wow, thanks for posting that. I switched to Ente for my photos recently, had no idea they also have a 2FA app. I was looking for a replacement for Aegis (after a switch to iOS), and this can even import from Aegis backup files. Neat. This means I can finally ditch my old phone I still had to have around just for 2FA :)
I love Ente Auth, but Ente (as a company/organization) does a somewhat poor job of calling out their non-photos apps in their branding and on their website. If you go to the "Download" button at the top of the page on this page about their LLM chat app, it downloads... their photo sharing application. If you click Sign Up, it takes you to a signup page with the browser title "Ente Photos" but the page text says "Private backups for your memories" with a picture of a lock - is that the Ente Auth signup, or the Ente Photos app signup?
A little bit of cleanup on their site to break out "Ente, our original photo sharing app" from the rest of their apps would do wonders, because I had to search around on the announcement to find the download for this app, which feels about like trying to find the popular Ente Auth app on their website
I just tried it. It downloaded Qwen3.5 2B on my phone and it's pretty coherent in its sentences, but really annoying with the amount of Ente products mentioned in every occasion.
Other than that it's fast enough to talk to and definitely an easy way to run a model locally on your phone.
As someone who saw this and was interested but also skeptical of this being low effort are there other open projects for running small models locally on android / iOS?
The "What's next" section is more interesting than what shipped. A general-purpose chat wrapper around a 1-4B model occupies a crowded space — PocketPal, Jan, LMStudio, GPT4All all do similar things. But the ideas they gesture at (a persistent "second brain" note, an LLM-backed launcher, long-term memory that grows with you) are actually differentiated
Have you tried WebLLM? Or this wrapper: CodexLocal.com
Basically, you would have a rather simple but capable LLM right in your browser using WebLLM and GPU
> This is not the beginning, nor is this the end. This is just a checkpoint.
Come onnnnnn. I would rather read a one line "Check out our offline llm" rather than a whole press release of slop.
This looks very neat. I'm not familiar with the nitty gritty of AI so I really don't understand how it can reply so quickly running on an iPhone 16. But I'm not even going to bother searching for details because I don't want to read slop.
Please god stop letting LLMs write your copy. My brain just slides right over this slop. Perhaps you have a useful product but christ almighty I cannot countenance this boring machine generated text.
This looks amazing! As I learn and experiment more with local LLMs, I'm becoming more of a fan of local/offline LLMs. I believe there's a huge gap between local LLM based apps and commercial models like Claude/ChatGPT. Excited to see more apps leveraging local LLMs.
The essence works, I was able to let it make a simple summary on CMS content. So next is making it do something useful, and making it clear how other plugins could use it.
It seems your link about the Wordpress variation validated my idea :).
If the new Wordpress feature would allow for connecting to Ollama, then there is no need anymore for my plugin. But I don't see that in the current documentation.
So for now, I see my solution being superior for anyone who doesn't have a paid subscription, but has a decent laptop, that would like to use an LLM 'for free' (apart from power usage) with 100% privacy on their website.
For when wordpress doesn't have enough exploits and bugs as it is. Also why bother with wordpress in the first place if you're already having an LLM spit out content for you ?
What's your point? Don't use LLM for CMS content? That my code is buggy? Or that people shouldn't trust the LLM they run on their computer on their own website?
You can check the code for exploits yourself. And other than that it's just your LLM talking to your own website.
> Also why bother with wordpress in the first place
Weird question, but sure, I use WordPress, because I have a website that I want to run with a simple CMS that can also run my custom Wordpress plugins.
I installed it and it's none of that. It is a mere wrapper around small local LLM models. And, it's not even multi-modal! Anyone could've one-shotted this in Claude in an hour (I'm not exaggerating).
What's the target audience here? Your average person doesn't care about the privacy value proposition (at least not by severely sacrificing chat model's quality). And users who do want that control can already install LMStudio/Llama.cpp (which is dead simple to setup).
The actual release product should've been what's described in "What's next" section.
> Instead of general chat, we shape Ensu to have a more specialized interface, say like a single, never-ending note you keep writing on, while the LLM offers suggestions, critiques, reminders, context, alternatives, viewpoints, quotes. A second brain, if you will.
> A more utilitarian take, say like an Android Launcher, where the LLM is an implementation detail behind an existing interaction that people are already used to.
> Your agent, running on your phone. No setup, no management, no manual backups. An LLM that grows with you, remembers you, your choices, manages your tasks, and has long-term memory and personality.
I think they did. If you start the download and then open the sidebar and/or background the app, the download progress bar disappears and is replaced by the download button. If you press the download button again, the progress bar reappears at the correct point.
I find that Claude often makes little statefulness mistakes like that. Human developers do too, but the slower and more iterative nature of human development makes it more likely that that would get caught.
This probably could have been one-shotted with Sonnet, not even Opus. Given how over indexed they are on LLM coding, Haiku might even be able to do it.
This is actually an interesting coding model benchmark task now that I think about it.
There is truly nothing original here and the product doesn't have a chance in hell of earning money. Local LLMs on-device will be dominated by the device vendors, whose control of the hardware stack combined with their ability to subsidize billions of dollars of machine learning research gives them an unfair advantage. Apple knows what the next generation of silicon will deliver, and their ML engineers are already hard at work building models that will be highly optimized for that silicon a year or two ahead of time. Open source models are really great and are backed by well funded labs; however, delivering these models on-device in a way that pleases users will never be easier than it is for the vendors of the devices.
Plus, device vendors have ways of making money from local LLMs that third-party app providers do not. They can make their local LLM free and earn money on the hardware play, without skipping a beat on the billions of dollars of ongoing R&D. I don't see how third party app vendors make money here when they will be competing with the decent, totally free alternative that Apple and Google (and Samsung etc.) will load on in the next year or two.
We have not seen a tidal wave of untechnical people vibe coding up their own software solutions.
When my little brother who is a drummer, and has never even looked at "code" before, had claude on-shot a python app that let him download songs on youtube, extract the stems, collect tempo/key/etc information, then feed that into his DAW, all without ever looking at code, knowing what any of it did, etc., I knew that we were about to see a LOT of single-use applications.
I'm not against it, honestly. I have always written little one-off scripts and apps that accomplished something faster than manually, now that those one-shots are possible with an LLM in seconds sometimes, it makes all my personal scripts so much easier... that said, I definitely read the scripts that are output, and attempt to step through them in a debugger before assuming it is all good.
That to me is more valuable than code vibe coded by Claude in one afternoon.
Ideally if you "participate" in the network, you would get "credits" to use it proportionally to how much GPU power you have provided to the network. Or if you can't, then buy credits (payment would be distributed as credits to other participants).
That way we could build huge LLMs that area really open and are not owned by any network.
I would LOVE to participate in building that as well.
This was posted the other day, but only briefly made the front page - seems kinda like what you’re talking about
Either LFM2.5-1.6B-4bit or Qwen3.5-2B-8bit or Qwen3.5-4B-4bit
Though, I don't see any references to Gemma at all in the open source code...
I would really like to know what people use these small and tiny models for. If any high-karma users are reading it, would you consider posting Ask HN?
very limited amount of use cases, perhaps. As a generalized chat assistant? I'm not sure you'd be able to get anything of value out from them, but happy to be proven otherwise. I have all of those locally already, without fine-tuning, what use case could I try right now where any of those are "very effective"?
Claude Code is a Desktop app as well.
For the user it's just important that the small grimlin that sits in the Ente app is not as smart as the grimlin that sits in the Claude app.
> Use Claude Code where you work
> Desktop Termianl IDE WEb and iOS Slack
Not that it is important any way ¯\_(ツ)_/¯
Have a comparison chart to Ollama, LMStudio, LocalAI, Exo, Jan.AI, GPT4ALL, PocketPal, etc.
However, it’s a bit confusing because, for example, a larger LLM model was downloaded to my smartphone than to my computer. It would probably make the most sense if the app simply categorized devices into five different tiers and then, depending on which performance tier a device falls into, downloaded the appropriate model and simply informed the user of the performance tier. Over time, it would then be possible to periodically replace the LLM for each tier with better ones, or to redefine the device performance tiers based on hardware advancements.
Going to give this a try...
Does this seem sound?
Here’s where it was added to PrivacyGuides - https://github.com/privacyguides/privacyguides.org/issues/36.... The person opening the issue is the CEO of ente. So the CEO of ente gets his company mentioned in PrivacyGuides back when it was new and that makes it more legit?
this seems self-contradictory
They just store tokens, without other FA at "worst" you get locked of your account but nobody else has access either. You're also supposed to, as good practice, not be limited to token generation and typically have a dozen or so of recovery tokens. Also if they were somewhat not working at doing the 1 task they should do, namely generate tokens, then you won't be able to use them so it won't even be added.
So... I might be missing something, can you please explain what worries you and why I should thus worry too?
https://en.wikipedia.org/wiki/Comparison_of_OTP_applications
So you look down you see a tortoise. It's crawling towards you.
What I'm missing is a way to create and use Passkeys across devices. My use case does not support creating a new Passkey on every device, I need to sync them via servers I control. The system that supports that will be the system that I migrate to.
Expressly harvesting creds through a 2FA app seems a little more direct.
How does it compare to Jan AI for example? or LM Studio? or ????
A little bit of cleanup on their site to break out "Ente, our original photo sharing app" from the rest of their apps would do wonders, because I had to search around on the announcement to find the download for this app, which feels about like trying to find the popular Ente Auth app on their website
https://github.com/ente-io/ente/blob/f254af939ff6950b63edf5f... Here is the system prompt, kinda embarassing
Helping non-technical people get off of ChatGPT.com and using increasingly better local models seems worth celebrating and continued iteration.
Absolutely no one called them crazy.
I've found https://github.com/alichherawalla/off-grid-mobile-ai but haven't tried anything in this space yet.
https://github.com/Arthur-Ficial/apfel
Apple Ai on the command line
https://github.com/alichherawalla/off-grid-mobile-ai
Then moved to pocket pal now for local llm.
Come onnnnnn. I would rather read a one line "Check out our offline llm" rather than a whole press release of slop.
This looks very neat. I'm not familiar with the nitty gritty of AI so I really don't understand how it can reply so quickly running on an iPhone 16. But I'm not even going to bother searching for details because I don't want to read slop.
I have a phone in a drawer I could install termux and ollama on over tailscale and then I'd have an always on llm for super light tasks.
I do really long for a private chat bot but I simply don't have access to the hardware required. Sadly I think it's going to be years to get there..
If Ente is reading this : please add requirements to make it run (how many RAM, etc.)
It requires a Firefox add-on to act as a bridge: https://addons.mozilla.org/en-US/firefox/addon/ai-s-that-hel...
There is honestly not much to test just yet, but feel free to check it out here, provide feedback on the idea: https://codeberg.org/Helpalot/ais-that-helpalot
The essence works, I was able to let it make a simple summary on CMS content. So next is making it do something useful, and making it clear how other plugins could use it.
Also: "Your AI agent can now create, edit, and manage content on WordPress.com" https://wordpress.com/blog/2026/03/20/ai-agent-manage-conten...
I'm talking about connecting Ollama to your wordpress.
Not via MCP or something that's complicated for a relatively normal user. But thanks for the link.
If the new Wordpress feature would allow for connecting to Ollama, then there is no need anymore for my plugin. But I don't see that in the current documentation.
So for now, I see my solution being superior for anyone who doesn't have a paid subscription, but has a decent laptop, that would like to use an LLM 'for free' (apart from power usage) with 100% privacy on their website.
For when wordpress doesn't have enough exploits and bugs as it is. Also why bother with wordpress in the first place if you're already having an LLM spit out content for you ?
You can check the code for exploits yourself. And other than that it's just your LLM talking to your own website.
> Also why bother with wordpress in the first place
Weird question, but sure, I use WordPress, because I have a website that I want to run with a simple CMS that can also run my custom Wordpress plugins.