Update 03 August, 2023

Thanks to everyone who’s participated in the DataPaymentNet testnet. It has performed its role magnificently, telling us that the fundamentals of data payments are sound, even if the UX leaves a little to be desired still. We’ve lost a few nodes after a replication spike the community spotted last night. But we’ve still 1675 MaidSafe nodes happily trucking along (and pre-spike we were at 2001 nodes for much longer than prior).

We’ve taken note of the errors: zero chunks after several hours, small clients running out of resources when self encrypting large files, Network error could not retrieve the record after storing it, ConnectionClosed, Did not store all chunks of file, OutgoingConnectionError to PeerId and others. Some of these are related: zero chunks is probably something to do with OutgoingConnectionError and UnroutablePeer, where the node is not being added to peers’ routing tables. Others will be the result of changes to communications functionality elsewhere, and some will be bugs that were previously covered up but have now emerged from their holes as we’ve fixed issues. But don’t worry - we’re ploughing through them all. :muscle:

This week, following a question on this forum @Qi_ma explains a little about Records.

Records and how they relate to Chunks, Registers and DBCs.

Record is a libp2p feature that enables data upload to the network and retrieval from the network as required.

The Record encompasses various essential functions at the network level, including tasks such as identifying the closest peers and facilitating data transfer. By delegating these network-level responsibilities to the Record, we can concentrate more effectively on data payment, transfer validation, security check, etc.

As part of our current implementation, we have distinct data types based on specific user cases. Specifically, we have Chunk to handle immutable data, DBC to manage spends and Register for Conflict-Free Replicated Data Types (CRDTs). These data types represent the information intended to be shared among the participants of the Safe Network. All of these data types can be encapsulated within the Record structure, thus enabling us to leverage its powerful functionalities.

By employing the Record, we can effectively store and exchange our custom data types, thus fostering seamless data sharing within the Safe Network.

Related to this is RecordStore which is a libp2p module designed to facilitate efficient storage of records onto disk while also serving as an interface to enable the execution of customised data tasks, such as data payment processing, transfer validation, security checks, and others. The RecordStore effectively bridges the gap between these custom data operations and the native libp2p Record, enabling seamless interoperability and integration within the system.

General progress

@Bzee has been looking at the UnroutablePeer and OutgoingConnection errors, which are rare issues, but not rare enough to ignore and seem to be related to nodes being ignored by other peers. They may be the result of recent fixes by the libp2p team, so he is conferring with them. The AutoNAT NAT detection function is another place he and @Aed900 are looking.

@Joshuef is also digging in here, as well as into ‘RecordParsingFailed’ and the occasional memory spikes reported by @shu and other community members. :pray:

@Qi_ma is leading investigations where our test setup can trigger a double spend on occasion, halting CI.

Meanwhile, on the internal testing front, @roland has fixed an infinite loop in the data verification process and @chriso continues to smooth the path to automated testnet deployment.

There have been discussions on how we can register as having a fixed address to ease several operations. Registers are mutable so their hash cannot be used as a network address, unlike, say, a Chunk. But there are ways around this, which @bochaco is spearheading.

@Anselme is preparing the ground, refactoring registers to remove type tags which are no longer needed

And looking a little further out @dirvine has been thinking about integrating SimpleX for DBC transfers and also the potential for local personal AIs to sit on the client and manage things like data storage, wallets and text-to-speech/speech-to-text for the user. There are lots of exciting things coming together in that area now, and Safe is in a great place to take advantage.

Useful Links

Feel free to reply below with links to translations of this dev update and moderators will add them here:

:russia: Russian ; :germany: German ; :spain: Spanish ; :france: French; :bulgaria: Bulgarian

As an open source project, we’re always looking for feedback, comments and community contributions - so don’t be shy, join in and let’s create the Safe Network together!


First now to rest this beauty


Second djdjdjjfjf


Just plopped in with a big Thurrrrrrrrrrd!!


Super update, team Maid! Looks like everyone is highly focused on specific tasks, which feels really positive in terms of being closer than ever to launch. Very exciting :fireworks: :clap:

Thanks to all for the hard efforts this week.

Cheers :beers:


SimpleX Chat: New Frontiers (Livestream from July 27, 2023)

How does it work without user IDs? 01:15.
Demo of the desktop app – you can already download beta version for Mac and Linux from GitHub. 03:51.
Setting up SMP relay with dual address (public+onion). 11:25.
SimpleX Chat funding. 18:23.
Future of SimpleX group and communities. 25:20.
Demo of SimpleX Directory Service prototype. 28:51.
Q&A. 28:51.

This is a recent presentation from the founder of SimpleX.


Thanks so much to the entire Maidsafe team for all of your hard work! :horse_racing:

And also for these wonderful testnets that give us hope! :horse_racing:


Thx 4 all the hard work Maidsafe devs

Great progress and nice read up, really valuable feedback from the community :clap: :clap: :clap: :beers:

Local personal AIs is a beautiful symphony in combination with the SAFE Network :exploding_head:

Keep hacking super ants


Fascinating Captain, as Spock would say. So now Warp 7 please, onward… Great update team!


@dirvine So SimpleX will be like a messaging/transfer protocol for DBCs, also chat/private messaging? Guess I’m kind of curious as well that if it wasn’t SimpleX then Maidsafe would have to have a transfer system, which I thought was just clients writing to the spentbook but from what I had gathered from previous conversation is SimpleX allows transfers to be private but the token to be more open/auditable or less private. I think there was mention of TPS benefits stripping the bulletproofing etc, but one time keys remain at least.

On the topic of client AI, is there some open library in mind? Is it something that is trainable on an individual basis? My minds doing laps wondering what the application of this will look like.

Can’t help but to think of what @JimCollinson or @oetyng (whom I haven’t heard from as much lately) are up to.

With these testnets doing so well, the pace is exhilarating so I understand it’s hard to stop and explain all of this. Great work by Maidsafe and the community.


Thinking here is speed to market, the focus of the team. There is also a real decentralisation move. Too.
MaidSafe → Data security and large scale fast payments
SimpleX (or any other) → communications

They are different networks which I kinda like as it adds that extra element of us not being all things to all folks but providing something stronger. Like detaching comms from data which feels right.

The hassle with all of this is usually interop and this then links to the work we are doing with personal secure AI models. The idea here is you won’t really have client apps as such (or not require them) and there would be a human API, a chatbot-like thing or indeed a speech-to-text and text-to-speech mechanism.

Given that then we can mix a whole bunch of apps/APIs together and the interface (AI) is just using all of those, wallets, data, comms etc. and you give it commands in your natural language.

I am hoping this allows anyone to use SAFE and even folk who cannot CLI/GUI with computers. It does not prevent folk from having their apps and using them like normal, but it means we can just shove APIs into the mix and the AI interface will be what merges the apps/networks together.

App devs then may just be prompt sharer folk. so the creatives can create amazing “apps” by chatting with their personal AI and sharing the apps they build with the AIs being what distributes and integrates them.

New world and some creative thoughts right now. I am interested to see where it takes us


I would like to understand better: the spendbook alone is not enough for transfer?

We can now pay for the data, that needs a transfer, right? What is missing then?


Good point

The SpentBook does not have the full output DBCs in it, just the identifiers. We could do a few tricks there and include the whole output DBCs perhaps. That would mean folks can just grab than and as only the recipient can Spend them it would perhaps be quite efficient.

Right now though we store them locally for transmission.


So at the moment when we ‘pay for data’ there is actually no counterparty receiving anything? We just… sacrifice? :grin:


We don’t really sacrifice. The payer has the DBC And must send it to the recipient.

We are discussing in house but likely we will make the whole output tDBC available and the receiver can collect it. Also though the sender can send out of band if they like as well. For nodes though collecting payments would be done easier if we actually send the DBC To them when the payment is made.

Nodes can be given the DBC on store as well. That also makes sense.


Is the intention for the AIUI to be a MaidSafe product, in which case who would be working on this and when (ie pre launch)? Also, when can I play? :smile:

Does this replace Jim’s Safe App or would that remain, like CLI etc for users who can’t run an AI (eg on their phone)?

Request: let’s not give this AI a cuddly name like Suzi etc, but maybe something non human, non cuddly that helps minimise anthropomorphizing.

Say, something like Armageddon :wink:

I think I’m going to call it that!


We will obvs need a topic for discussion about Armageddon, sorry I mean the SN User Agent :wink: but is it credible as a general UI for consequential tasks?

We know LLMs make errors and errors in Safe Network are potentially more consequential (eg irreversible) than in the systems SN aims to replace such as for money transfers, backing up all your family photos, sending very private information to the correct recipient etc.

Would you trust an AI with a speech or text interface with more than say drafting a tweet?

What kind of actions would you trust it to carry out?

What kind of actions would you NOT trust it with?

Seems worth exploring those questions before developing this.


Speed to market and more decentralized is good. SimpleX seems very, well simple! Much like Maidsafe designs strive to be and I really like that. I do also like that data and comms are wrapped up as a full offering though too. Anymore one is less without the other.

So I’m curious is this a potential partnership? Will they be rolled into one so when I download the SN client I get ALL the features. My assumption here is yes but feel I should still clarify.

The personal client AI sounds really cool. I’d still like the option and granularity of Jim’s UI too but the client AI to get straight to market (assuming that is faster) would be great, at first.
Maybe my opinion would change but again, I think having the granularity and agency with the Safe App UI is still a high priority to me. Just after launch and AI client if that shortens time to launch.

Loving how quickly things are coming along now and the grit and determination to launch this as soon as possible while remaining up to date or even ahead of the curve. That’s something enticing to build on.


What news! The update sounds excellent as testnet is stable and more problems have been discovered.

Many thanks to the team and the testers for a fantastic job, as the problems found are not trivial, and you can also see that the community’s commitment to testnet is growing! :clap: :blush:

When you posted the thread about SipmleX I was sure you had a combination of SimplyX and DBC in mind :grin: however the combination of the two with personal secure AI models looks like a small hydrogen bomb :bomb: :point_left: :blush:
@happybeing very aptly called it Armageddon :wink:

I have not commented on AI in the forum yet because it is a huge topic, but it is probably worth considering some of the implications of such an Armageddon :blush:

If AI were just a tool to assist humans rather than replace them, as is the current trend, this could be an excellent solution, but the question that comes to mind at the moment is whether this would complicate or delay the implementation of the basic functionality of SafeNet?

On the other hand, in principle, the role of AI comes into play when we have huge data resources that can be used with AI to achieve greater benefits, whereas I associate the AI interface assistant with something that is supposed to decide for me…
Are my concerns justified?

What worries me most, however, is the idea of using a chatbot that could become SafeNet’s weakest link - easily eavesdropping on the user and recognising them by their voice ‘ear’, and how it is powered by AI…