We've used Element with my inlaws for a few years now. Most of the time it's fine but they really need to fix those issues with memebers of a group loosing the encryption keys and have more transparent feedback channels. Not sure if they finally fixed the fact that uploads for high res videos would always fail
Within my friend group, we also switched now to Matrix. One of us is hosting it on their server, and we also use ElementX as clients. We did this to try get away from lack of privacy on Whatsapp and Telegram. It was also fun to tinker around with a selfhosted solution.
In particular I love that I can now share my live location with this friendgroup and be very confident that noone is harvesting this data. With whatsapp, I was always very uneasy allowing live location access.
Apart from one friend having issues receiving notifications, overall, we are having a positive experiene on ElementX/Matrix. I like especially how clean and well designed the iOS App looks. It simple and it just works.
If you stop logging in regularly to the other ones, people eventually move over when they don't want to wait a day or two for a response. Just make it clear you're only on matrix 24/7, anything else you're just not active on. (Which means uninstall the mobile apps, too.)
Also, Discord's id scan rollout was extremely effective in motivating my friend group. I don't think any of them even got directly hit by it (yet), but we've all had other issues with their faulty AI bans and account flags and losing accounts due to them that it wasn't a tough sell for us.
My online friend group made the switch earlier this year when Discord was really pushing on the ID verification thing.
We just made detailed instructions and stuck to locking the discord server by a certain date. Everyone who cared to interact moved over either then or shortly after when they noticed that the server was locked. I think this mainly worked out because we were willing to let go of anyone that would neither speak against switching nor make the switch.
If they won’t switch they aren’t serious people and they are better off staying elsewhere. My community lost more than half of its members and we have been using Matrix for over a year.
Yes and if someone is not a commoner but a journalist investigating corruption in your city he can be - as has happened to many journalist worldwide - targeted because you only focus on immediate first order effects to you.
Can you be sure about that? What if you purchased things and voted for politicians because of well placed targeted content?
We all feel like we're in control of our decisions, but if advertisers are paying for ads and seeing increased sales, some of us must be buying things.
There are three technologies I can name which are making me more concerned with privacy and data collection:
1. AI-powered "personalized pricing" for obvious reasons.
2. Advertising. Effective advertising captures my attention (a scarce resource) and "creates demand" in my brain for products I otherwise wouldn't need. No TV ad has ever led to me purchasing something, but Instagram ads and YouTube affiliate marketing fuel my B.A.S. (bike acquisition syndrome).
3. Predatory apps like sports gambling and MMORPGs. Like advertising, these are designed to create addiction and using personalized data makes it all the more effective.
While I am also a commoner with avg habits, I would like to protect myself against sophisticated attackers like Instagram Ads, RuneScape, DraftKings, and predatory pricing at the grocery store. Giving up my privacy and data gives the attacker an advantage (they know how to manipulate me).
None of this has anything to do with government surveillance. We all hope for philosopher kings who wield Palantier with grace and dignity, but history has more examples of despots using technology to oppress us commoners :(
Same but with XMPP! Much easier to host (there is an order of magnitude difference in resource consumption). Although my client can't do live location, only manually sending the current point.
Why is the resource consumption of the Matrix server implementations so high? What are they doing that hits so heavy on performance? Encryption? But isn't it e2e? State resolution? Surely the resolution algorithms can't be that bad?
Even Dendrite, the eternally-WIP-rewrite in Go, used so much RAM I couldn't even start it on my Raspi.
The reason Matrix typically consumes more resources than XMPP is because it's replicating a large DAG of state events (i.e. the key-value data associated with a room; not the room's actual messages) between all participating servers. So, in a room with 100K users, each server has to maintain at least 100K state events (every time someone joined a room, or changed their name, or left the room, or set permissions, or changed the room's name etc).
Then, whenever the state of the chatroom changes (e.g. you join it, or someone else joins it), each server has to resolve its view of the state of the room with the proposed change. We call this merge conflict resolution algorithm 'state resolution', and it's typically where the heaviness comes from, as a naive implementation loads up all the state events for the room in order to calculate and store the new state.
In exchange, you get fully decentralised access control, and an eventually consistent zero-trust distributed key-value store, with resilience to network partitions (including in P2P environments) - unlike (say) vanilla XMPP which applies access control and KV on a single given server per-conversation.
Now, the key thing is that *there is nothing in the protocol which makes this intrinsically heavy to implement*. The only reason that servers can be heavyweight today is simply because more time has been invested in chasing down more pressing issues like encryption issues, investing in client SDKs, speeding up sync responses, switching to OIDC for auth, etc.
The problem is made worse by the fact that Synapse (the original Python server) has historically been particularly clunky, skewing perception. Over the last few years this has improved significantly, although it could still be way better - at the Matrix Conference last month (https://youtu.be/kdEDxI2MCjA?list=PLFDUAtPqCX98&t=329) we showed off a fork which had improved CPU 10x by shifting the hot path entirely to Rust from Python. (It turns out that over 2026, we wrote as much Rust for Synapse as we did Python, ignoring tests). Further improvements would be to memoize the state res and calculate it incrementally rather than starting from scratch every time. Similarly, there are 100x improvements to be had on storage usage - e.g. by storing state temporally (c.f. my https://docs.google.com/presentation/d/1pKtLl4vCV3-8xz8crvxW... spike from last year).
So, why haven't these improvements landed yet? In practice, the team at Element who builds Synapse have had to focus on customer-driven work instead in order to keep the lights on, and frustratingly performance is lower down the todo list than other feature-work. On the plus side, over the last year or so Element has got into a much healthier state - hence us being able to show off the 10x improvements in Synapse last month (although they still need to land).
Meanwhile, elsewhere in Matrix there are other servers which are more actively focused on being lightweight right now (e.g. continuwuity), but don't have the maturity of Synapse (or the commercially supported distribution that Element provides). And even then, there's still a lot of headroom for speeding up and shrinking things - it should be very very doable to run Matrix on a tiny host like an RPi.
It's not a perfect analogy but I think of it as being similar to the evolution of DVCS systems. There's nothing intrinsically heavy or hard about replicating DAGs of source code between hosts, but Monotone was pretty slow and clunky while Git showed that a smarter implementation could be more than performant enough. The race is on to build the equivalent for Matrix.
Signal is a non profit with robust terms that forbid monetisation of data.
The data you share is minimal anyway, and can’t be used to decrypt any messages.
It’s not suitable for any industries that require any form of record keeping or auditing, the minimal tradeoffs are well worth it, given that you don’t have to stand up your own solution.
Some people (like me) simply don't like the idea of being at the mercy of any single particular entity. No amount of convenience or practicality is going to convince those people that Signal is worth adopting. If I am going to go through the hassle of getting my friends and contacts out of WhatsApp, then I'd rather just skip Signal and go straight to a protocol that supports multiple clients and does not have its preconceived notion about how people should communicate.
There is a technical guarantee with Signal, and their ability to cooperate with governments has already been tested, when the FBI came to Signal with a court order for all information they had on a given user, and Signal turned over everything they had: the user's sign up date.
For years now, Signal has uploaded people's entire contact lists to their server, protected only by Intel SGX that is infamously vulnerable. I use Signal myself, and suppose that its chats themselves may be reliably E2EE. But it's a reasonable assumption that a nation-level actor has access to contact metadata, not just the bare facts Signal says it discloses when authorities come calling.
Connect with other matrix servers? As far as I understand, this is the hard part and makes scaling difficult. (One of a servers user joining a group on a different server will make its original server load all that group data as well and that sums up)
For your use case I also considered setting up a instance. Connecting a cluster of friends. But it would be nice if it could grow a little bit, without problems and the server choking.
This is a disaster. Matrix's encryption is a tire fire. They continually say they're going to fix it but it's still no where near the security and privacy needed for national security use-cases.
I recently listened to a podcast addressing this (one of the hosts works on Matrix/Element), basically, the "Unable to decrypt message" is what almost any error ends in. Also, they recently tried to remove all these issues and managed a 10x reduction, another highly focused effort may reduce it another 10x. I believe this was on the Late Night Linux podcast [0].
In any case, I use Matrix for anything FOSDEM related and I have seen 1 of these errors over the last year, said user said they replied from within an Android notification fwiw.
> basically, the "Unable to decrypt message" is what almost any error ends in
That's not the issue(s) I'm referring to. And typical of Matrix's lack of engagement with weaknesses and issues that have been raised several times. They continue to neglect the issues their protocol has.
I wanted to see what are those issues but I do not want to read the whole transcript especially written white on black which strains the eyes so I had to search for the right place. Podcasts are an awful thing it seems.
There's a lot of parallel work going on to improve encryption in Matrix, including MLS. That particular effort is paused, but there are various other efforts.
You also have to remember that this is a million times better than the status quo: Teams. Cleartext directly to the US.
Having said that if they were upfront about the shortcomings of the current protocol, I doubt some agencies in Europe would have chosen Matrix. Then there’s the promise to add MLS at some point. My guess is that’s years away: in the Q&A of one of their talks, they said they don’t yet know how to make MLS work with a decentralized architecture.
As of now you seem to suggest that lack of e2ee outweighs probable safety and security because there's little chance of inviting a third party to the chat.
Pretty sure EC staff will not be allowed to use it for highly sensitive communication. You will not see von der Leyen use this. All the rest can just add to the transparency of governance that we all want right?
> Live presence was switched off. As more users came on board, the load from live presence (the real-time “online/away” indicator) started to affect performance, so the team disabled it.
im taken back to my AIM days; imagining someone setting up special alert sounds for when their favorite official comes online
This seems surprising to me. I think live presence is an essential part of real time communication, but also something that should be well understood how to scale?
I don't understand why being asynchronous would mean you don't want to know status, in fact for me it's the opposite. With purely synchronous real-time, e.g. plain old telephone service (POTS), you don't need status, you just dial their number and see if they pick up.
It's async where status helps a lot, particularly if it's enrichable with "Back in an hour" or "Gone for the weekend", or "Idle: 2hr", or "Currently in Meeting", or whatever.
i could see how after 10k users, the logarithmically? growing time/resource pressure of checking more users against every list, without properly distributing the data / sharding, starts to creep in
Perhaps I'm not understanding the Matrix model, is there no central broker that is holding user contact relations and fanning out and directing statuses / messages?
Because, otherwise, even if you went crazy and did the most naive implementation:
Storing for every User
User -> List <User Contacts>
And the reverse:
User -> List<Users who have you as contact>
And even then if you go full quadratic and have literally everyone have everyone else as a contact, and store and update the online status for each pair, and do that for each direction, you still shouldn't have scaling issues at 50k users, as that just isn't that much data?
By 500k Users you might be starting to feel it with a naive approach, but 50k? It shouldn't be a bottleneck unless status updates are causing cascade effects into other systems.
And that's just an analysis of the most naive design I could think of!
based on experience with another tool i saw, each lookup itself must be authorized by checking membership in at least two other lists on three other servers :P
Indeed, this seems like a really low number for when to run into issues, unless they're running it on a shared VPS and/or it's just not very well built.
I mean, if the user names and the messages themselves are cached and displayed, why can't user location be done the same way every ~5-10min or so, or ~1m with an active direct user-to-user (with approval) lookup? it's an exchange of two ~6-10 digit grids and an elevation, I'll give allowance for a timestamp.
It's not about better, its about incentives. Someone will be incentivised to move to a paid platform as it will look better on their CV (I sucessfully implemented a major 15 million euro project), and Slack will of course help them with the finance case
It's the public sector. No one is incentivised to save any money. In fact, saving money may make someone think it's okay to reduce your budget, which is bad.
Someone who saves the government money by decoupling it from the private sector is probably not going find many friends in said sector after their tenure ends. Revolving doors, like any machine, needs greasin'...
Someone will have done, then they move on, and the next owner needs to make their mark, meanwhile the sales people are explaining how much better their proprietary system is over nice lunches, people who come in who are used to slack/teams/etc are whining that this is different and therefore worse, and someone says "where's the piece of paper which stops me being held accountable if it breaks"
I've found Matrix really nice for isolated setups, funnily enough.
E.g. my wireguard network has an isolated/non-federated Matrix server all my AI agents to talk to me over. It's great for that, but I have no interest in using it to talk to people.
We tried it out for a few weeks earlier this year with a small team. It's alright but there are some annoying defaults (no link previews by default being one of the more annoying ones). You have to opt in to that on every channel for every user. I get why they did that but basically usability and UX is thrown under the bus here because of some hypothetical privacy issues.
That's nice if you are genuinely paranoid on this front (rightfully or otherwise) but really sucks if you want people to share links with each other (one of the big use cases we seem to have for using tools like this).
Related to that, a side effect of all the e2e encryption is that it's really hard to give people or agents access to the full history of a channel. It's not impossible but it involves jumping through some hoops. And of course a primary reason for us switching away from Slack was using AI agents in our chats more and experimenting with running some of our own bots and something that is sometimes referred to as a company brain.
Instead of fighting the tool we switched the tool. We picked Zulip around the same time the founders of that project were hired by Anthropic. For the migration, I gave AI access to my matrix keys and it moved over most of our channels to Zulip. Since Zulip has topics, I also made it group messages by topic.
I like Zulip so far. They could use a UX refresh in the client. It looks a bit dated and meh. But otherwise it does the job. With both Zulip and Matrix, I orchestrated the rollout and hosting with AI, including giving admin privileges to manage the whole thing. Works surprisingly well with both. But jobs like add this person to that channel and given them such and such rights are working flawlessly. We had some openclaw thing running for a while but we gave up on that (too much issues with that). But we have a lot of chatgpt scheduled automations and some scripts and skills that allow it to search through and post in Zulip channels. Very useful.
If you don't want to self host, Zulip has a generous freemium tier as well in their managed hosting. Haven't tried it but looks like a decent starting point if you want to try it out. You can migrate to self hosted later, if you want.
> no link previews by default being one of the more annoying ones)
It's not just about privacy. To me, this is more usable - it always annoys me, the default habit of taking up half a (phone) page with a giant site card. If it just resolved title (and used normal chat-sized font) it would probably be good, but nothing seems to do that.
I'm so sad XMPP never took off. I remember being annoyed by the fragmented state of support for pretty vital "extensions" though. Has anything changed in that regard? I remember there was talk of consolidating a bunch of must-have extensions, like E2EE, into a single highly recommended one – did that go anywhere?
I really really hope that matrix with fractured client landscape, session verification, encryption errors, half-baked spaces, usability problems and slow development (Element X when?) improves. I use it daily and the problems are real. EC switching to matrix might add some development forces and money.
If you actually stop logging in regularly to the other ones, people eventually move over when they don't want to wait a day or two for a response. Just make it clear you're only on matrix 24/7, anything else you're just not active on.
As you can see this blog is from "Element", who are offering Matrix deployments for enterprise and governments as a service, so they are definitely being paid for their work. As the company was founded by the Matrix creators, I'm sure they are also spending dev resources on it, but you can read more on that here: https://element.io/en/matrix-relationship
I personally think this is a good example of how you can build an open-source product and still profit of it commercially by offering it as a vendor to enterprise and governments.
Well, leave it to the bureaucrats in Brussels, to spend years, and several tens of millions (low estimate) in a decision and implementation that ends up deciding the best path to take was: Matrix.
Given how much time I spend "verifying my device" and other issues with simply making it work, and multiplying that by 40000, I feel very sorry for them.
Yes it is awful. I do not even understand how I can "verify" a temporary browser session which gets destroyed after a log out.
Reminds me of WhatsApp vs Telegram. In Telegram to log in from a new device, you just enter the phone number and the code. In WhatsApp you need to scan some codes, do other things, overcomplicated.
Long time enemy l element and matrix user here.
Indeed the verification process of new sessions is complicated.
Something they'll surely address as they grow (until no security advantage is left...)
> The European Commission's rollout of Element Pro as a self-hosted alternative to Microsoft Teams has received harsh criticism from officials, with users citing poor usability and calling the tool "absolute shit."
Teams is such a shitshow. I always join meetings by using an incognito browser, pasting the link, and staying in browser. If i try the app, or logged in browser it fails, without giving any useful error or message to work with.
I'm sure you will find more than enough users that would also describe Microsoft Teams as absolute shit, but it is also not intended to replace Teams, but offer an additional channel for more sensitive communication that you wouldn't want to have via Teams.
Somehow it's not surprising that the readership here doesn't understand why would Europe want to pay the high price of riding itself of US dependencies, but somehow it's still surprising.
My impression is that element is secure, Europe needs secure, but it does not imply that we need element. There are other secure solutions, but also enterprise element might be different from the consumer version that so many people dislike.
I mean good. But given how much impacts Russia has on some EU countries, how many people it can bribe to do its bidding. I am waiting for some leak here.
Europeans do not develop seem to be developing for the masses. The French were allegedly developing a google drive alternative, for French civil servants.
If it was for our institutions, the AI would only be available for civil servants.
We despise the American mass surveillance tools but we crave them and depend on them
There are dozens if not hundreds of Drive and office alternatives in all regions of the globe, designed to fill any number of niches, including data sovereignty.
They’re not as big as Google, but many of them have much better support and features.
You won’t see these players on the front page of HN but you’ll come across them if you work in or around government.
> The French were allegedly developing a google drive alternative, for French civil servants.
That seems a bit unfair. Most of the new tools developed for the civil service and government are open source and can be used by anyone [1]. The servers are not (it’s a combination of self-hosting and contracts with companies), but that’s the same anywhere and I don’t really see how it could be (or should be) different, considering the requirements.
It’s better than the alternative, which is to do exactly the same thing with proprietary software from mostly American companies (which itself would not be too much of a problem if it were not for things like the CLOUD act).
> If it was for our institutions, the AI would only be available for civil servants.
I don’t see what makes you think that. They are already deploying internal AI tools (“they” being very broad, it’s true of both member states governments and EU institutions) and it is not incompatible with companies providing services to the general public. It’s the same in the US: random people in Louisiana cannot use the NSA’s infrastructure or access internal tools at the DoD or the DoE’s HPC clusters.
In particular I love that I can now share my live location with this friendgroup and be very confident that noone is harvesting this data. With whatsapp, I was always very uneasy allowing live location access.
Apart from one friend having issues receiving notifications, overall, we are having a positive experiene on ElementX/Matrix. I like especially how clean and well designed the iOS App looks. It simple and it just works.
In my experience, most people just don't care enough
Also, Discord's id scan rollout was extremely effective in motivating my friend group. I don't think any of them even got directly hit by it (yet), but we've all had other issues with their faulty AI bans and account flags and losing accounts due to them that it wasn't a tough sell for us.
We just made detailed instructions and stuck to locking the discord server by a certain date. Everyone who cared to interact moved over either then or shortly after when they noticed that the server was locked. I think this mainly worked out because we were willing to let go of anyone that would neither speak against switching nor make the switch.
I'm just a commoner with avg habits that have seen 0 consequences in 30 years of live from the Evil Corps using my (already easily available) data
We all feel like we're in control of our decisions, but if advertisers are paying for ads and seeing increased sales, some of us must be buying things.
1. AI-powered "personalized pricing" for obvious reasons. 2. Advertising. Effective advertising captures my attention (a scarce resource) and "creates demand" in my brain for products I otherwise wouldn't need. No TV ad has ever led to me purchasing something, but Instagram ads and YouTube affiliate marketing fuel my B.A.S. (bike acquisition syndrome). 3. Predatory apps like sports gambling and MMORPGs. Like advertising, these are designed to create addiction and using personalized data makes it all the more effective.
While I am also a commoner with avg habits, I would like to protect myself against sophisticated attackers like Instagram Ads, RuneScape, DraftKings, and predatory pricing at the grocery store. Giving up my privacy and data gives the attacker an advantage (they know how to manipulate me).
None of this has anything to do with government surveillance. We all hope for philosopher kings who wield Palantier with grace and dignity, but history has more examples of despots using technology to oppress us commoners :(
Even Dendrite, the eternally-WIP-rewrite in Go, used so much RAM I couldn't even start it on my Raspi.
And yes the algorithms are tough - especially if you’re in the bigger rooms with lots of servers that don’t always work right.
Then, whenever the state of the chatroom changes (e.g. you join it, or someone else joins it), each server has to resolve its view of the state of the room with the proposed change. We call this merge conflict resolution algorithm 'state resolution', and it's typically where the heaviness comes from, as a naive implementation loads up all the state events for the room in order to calculate and store the new state.
In exchange, you get fully decentralised access control, and an eventually consistent zero-trust distributed key-value store, with resilience to network partitions (including in P2P environments) - unlike (say) vanilla XMPP which applies access control and KV on a single given server per-conversation.
Now, the key thing is that *there is nothing in the protocol which makes this intrinsically heavy to implement*. The only reason that servers can be heavyweight today is simply because more time has been invested in chasing down more pressing issues like encryption issues, investing in client SDKs, speeding up sync responses, switching to OIDC for auth, etc.
The problem is made worse by the fact that Synapse (the original Python server) has historically been particularly clunky, skewing perception. Over the last few years this has improved significantly, although it could still be way better - at the Matrix Conference last month (https://youtu.be/kdEDxI2MCjA?list=PLFDUAtPqCX98&t=329) we showed off a fork which had improved CPU 10x by shifting the hot path entirely to Rust from Python. (It turns out that over 2026, we wrote as much Rust for Synapse as we did Python, ignoring tests). Further improvements would be to memoize the state res and calculate it incrementally rather than starting from scratch every time. Similarly, there are 100x improvements to be had on storage usage - e.g. by storing state temporally (c.f. my https://docs.google.com/presentation/d/1pKtLl4vCV3-8xz8crvxW... spike from last year).
So, why haven't these improvements landed yet? In practice, the team at Element who builds Synapse have had to focus on customer-driven work instead in order to keep the lights on, and frustratingly performance is lower down the todo list than other feature-work. On the plus side, over the last year or so Element has got into a much healthier state - hence us being able to show off the 10x improvements in Synapse last month (although they still need to land).
Meanwhile, elsewhere in Matrix there are other servers which are more actively focused on being lightweight right now (e.g. continuwuity), but don't have the maturity of Synapse (or the commercially supported distribution that Element provides). And even then, there's still a lot of headroom for speeding up and shrinking things - it should be very very doable to run Matrix on a tiny host like an RPi.
It's not a perfect analogy but I think of it as being similar to the evolution of DVCS systems. There's nothing intrinsically heavy or hard about replicating DAGs of source code between hosts, but Monotone was pretty slow and clunky while Git showed that a smarter implementation could be more than performant enough. The race is on to build the equivalent for Matrix.
The data you share is minimal anyway, and can’t be used to decrypt any messages.
It’s not suitable for any industries that require any form of record keeping or auditing, the minimal tradeoffs are well worth it, given that you don’t have to stand up your own solution.
Plus lack of any org level management
Connect with other matrix servers? As far as I understand, this is the hard part and makes scaling difficult. (One of a servers user joining a group on a different server will make its original server load all that group data as well and that sums up)
For your use case I also considered setting up a instance. Connecting a cluster of friends. But it would be nice if it could grow a little bit, without problems and the server choking.
Not to mention that you can host chatmail on a potato
Their work to move to MLS seems to be abandoned. https://arewemlsyet.com/
I applaud the people working on Matrix, it worthy project, but I feel the E2EE issues are at a level close to negligence.
In any case, I use Matrix for anything FOSDEM related and I have seen 1 of these errors over the last year, said user said they replied from within an Android notification fwiw.
I really enjoy Matrix/Element tbh.
[0] https://latenightlinux.com/
That's not the issue(s) I'm referring to. And typical of Matrix's lack of engagement with weaknesses and issues that have been raised several times. They continue to neglect the issues their protocol has.
This podcast goes into some of the details: https://securitycryptographywhatever.com/2022/11/02/Matrix-w...
However, I would be prepared for a "we fixed those issues" claim which really amounts to shallow fixes that move the issue elsewhere.
The proposal to move to MLS was accepted for a lot of reasons, reasons that have not, or cannot be addressed with their current protocol.
You also have to remember that this is a million times better than the status quo: Teams. Cleartext directly to the US.
> You also have to remember that this is a million times better than the status quo: Teams. Cleartext directly to the US.
I would agree with this, but it's also being proposed to replace things like Signal.
https://cfp.2026.matrix.org/matrix-conference-2026/talk/G9XJ...
https://cfp.2026.matrix.org/matrix-conference-2026/talk/7GYB...
Re: Signal, there are more concerns than just encryption. For example: https://element.io/blog/how-element-protects-against-signalg...
Having said that if they were upfront about the shortcomings of the current protocol, I doubt some agencies in Europe would have chosen Matrix. Then there’s the promise to add MLS at some point. My guess is that’s years away: in the Q&A of one of their talks, they said they don’t yet know how to make MLS work with a decentralized architecture.
As of now you seem to suggest that lack of e2ee outweighs probable safety and security because there's little chance of inviting a third party to the chat.
im taken back to my AIM days; imagining someone setting up special alert sounds for when their favorite official comes online
Send me a message, I'll get back to it when I'm online.
It's async where status helps a lot, particularly if it's enrichable with "Back in an hour" or "Gone for the weekend", or "Idle: 2hr", or "Currently in Meeting", or whatever.
Because, otherwise, even if you went crazy and did the most naive implementation:
And the reverse: User -> List<Users who have you as contact>And even then if you go full quadratic and have literally everyone have everyone else as a contact, and store and update the online status for each pair, and do that for each direction, you still shouldn't have scaling issues at 50k users, as that just isn't that much data?
By 500k Users you might be starting to feel it with a naive approach, but 50k? It shouldn't be a bottleneck unless status updates are causing cascade effects into other systems.
And that's just an analysis of the most naive design I could think of!
I'd be willing to bet that they migrate to something else after 12 months.
E.g. my wireguard network has an isolated/non-federated Matrix server all my AI agents to talk to me over. It's great for that, but I have no interest in using it to talk to people.
That's nice if you are genuinely paranoid on this front (rightfully or otherwise) but really sucks if you want people to share links with each other (one of the big use cases we seem to have for using tools like this).
Related to that, a side effect of all the e2e encryption is that it's really hard to give people or agents access to the full history of a channel. It's not impossible but it involves jumping through some hoops. And of course a primary reason for us switching away from Slack was using AI agents in our chats more and experimenting with running some of our own bots and something that is sometimes referred to as a company brain.
Instead of fighting the tool we switched the tool. We picked Zulip around the same time the founders of that project were hired by Anthropic. For the migration, I gave AI access to my matrix keys and it moved over most of our channels to Zulip. Since Zulip has topics, I also made it group messages by topic.
I like Zulip so far. They could use a UX refresh in the client. It looks a bit dated and meh. But otherwise it does the job. With both Zulip and Matrix, I orchestrated the rollout and hosting with AI, including giving admin privileges to manage the whole thing. Works surprisingly well with both. But jobs like add this person to that channel and given them such and such rights are working flawlessly. We had some openclaw thing running for a while but we gave up on that (too much issues with that). But we have a lot of chatgpt scheduled automations and some scripts and skills that allow it to search through and post in Zulip channels. Very useful.
If you don't want to self host, Zulip has a generous freemium tier as well in their managed hosting. Haven't tried it but looks like a decent starting point if you want to try it out. You can migrate to self hosted later, if you want.
It's not just about privacy. To me, this is more usable - it always annoys me, the default habit of taking up half a (phone) page with a giant site card. If it just resolved title (and used normal chat-sized font) it would probably be good, but nothing seems to do that.
> https://www.politico.eu/article/eu-microsoft-teams-alternati...
> The U.S. would “instantly” win a hypothetical tech war, an official moaned.
This is the same system.
I personally think this is a good example of how you can build an open-source product and still profit of it commercially by offering it as a vendor to enterprise and governments.
https://en.wikipedia.org/wiki/Amdocs
https://en.wikipedia.org/wiki/Matrix_(protocol)#:~:text=The%...
Edit (downvotes within literally 3 seconds lol)
Reminds me of WhatsApp vs Telegram. In Telegram to log in from a new device, you just enter the phone number and the code. In WhatsApp you need to scan some codes, do other things, overcomplicated.
There's really only so many ways you can do that securely.
> The European Commission's rollout of Element Pro as a self-hosted alternative to Microsoft Teams has received harsh criticism from officials, with users citing poor usability and calling the tool "absolute shit."
It doesn't seems they're enjoying it
I'm astonished people can deal with using Teams.
They’re not as big as Google, but many of them have much better support and features.
You won’t see these players on the front page of HN but you’ll come across them if you work in or around government.
That seems a bit unfair. Most of the new tools developed for the civil service and government are open source and can be used by anyone [1]. The servers are not (it’s a combination of self-hosting and contracts with companies), but that’s the same anywhere and I don’t really see how it could be (or should be) different, considering the requirements.
It’s better than the alternative, which is to do exactly the same thing with proprietary software from mostly American companies (which itself would not be too much of a problem if it were not for things like the CLOUD act).
> If it was for our institutions, the AI would only be available for civil servants.
I don’t see what makes you think that. They are already deploying internal AI tools (“they” being very broad, it’s true of both member states governments and EU institutions) and it is not incompatible with companies providing services to the general public. It’s the same in the US: random people in Louisiana cannot use the NSA’s infrastructure or access internal tools at the DoD or the DoE’s HPC clusters.
[1] the Google Drive equivalent, for example, is here: https://github.com/suitenumerique/drive