Hi
anyone is aware of this strange boinc project : https://axiom.heliex.net/

IMPORTANT: Modified BOINC Client Required
The standard BOINC client is not fully compatible with Axiom streaming work units.
You must use our modified client for proper operation.
Download Axiom BOINC v2.42

"Distributed AI"... there is no useful information on the forum, only short messages (with no link) on the home page, no clear communication.

On the forum some crunchers are trying it, but all is strange here.

It seems it is creating a folder out of the boinc data structure and uses it to read some input files. I understand boinc security system is broken here.

See also messages on WUProp forum:
https://wuprop.statseb.fr/forum_thread.php?id=191#12098

---

Paul.

I Definitely will NOT download and install a custom/modified BOINC client for a project, especially if it stores data OUTSIDE the normal BOINC locations.

Who knows what else it might be attempting to read/write on the computer, your local network or Internet.

The deliberately vague (my description) in the about page of the project consists of gobbledegook about an endocrine system mimicking the effect of cortisol in the body. Until they can give a coherent account of what the project's aims are, I won't be running it, especially with their own customised client.

The standard BOINC client is not fully compatible with Axiom streaming work units.

I wouldn't touch it with a barge pole.
Sounds like a great way to have your system hijacked.

---

Grant
Darwin NT.

I wouldn't touch it with a barge pole.
Sounds like a great way to have your system hijacked.

I am toying with the idea of trying it in a VM to try and work out if it is doing anything dodgy.

The admin said they're moving back to the standard client and a wrapper.

PyHelix

— 2:23 AM
Alright I moved the change from the client to wrapper:
Implemented soft suspend in the wrapper as suggested
Uses a signal file for the app to acknowledge before being killed
Users can now use the standard BOINC client
The wrapper is generic and could benefit other projects with streaming apps
Source: https://github.com/PyHelix/boinc-Axiom (streaming_wrapper.cpp)

Now first page says

Getting Started
1. Download BOINC (standard client)
2. Add project: https://axiom.heliex.net
3. Place files in %USERPROFILE%\Axiom\contribute
The streaming wrapper handles suspend/resume automatically. Source code (GPL)

(... very clear...)

@mmonnin I don't see where you got these explanations, not from the forum ?

Far too much vagueness about the project objectives, and why it needs to use streaming - strikes me as a (personal) data harvesting device, thus I'm avoiding it like a plague.

In reply to JeromeC's message of 4 Feb 2026:

Now first page says

Getting Started
1. Download BOINC (standard client)
2. Add project: https://axiom.heliex.net
3. Place files in %USERPROFILE%\Axiom\contribute
The streaming wrapper handles suspend/resume automatically. Source code (GPL)

(... very clear...)

@mmonnin I don't see where you got these explanations, not from the forum ?

The only form of communication the admin has ever done is in Discord.
https://discord.com/channels/605398931479724074/605398932067188768

They now have Linux tasks that run on a standard client. Most of the processing time, top shows very little CPU activity even with 16 tasks running at once. Tasks are taking about 6 minutes on my Ryzen9. Probably a bit quicker if I wasn't running them in a VM.

@Dave : Did you have a look at network bandwidth consumption ? are these tasks "streaming" anything ?

And regarding the "3. Place files in %USERPROFILE%\Axiom\contribute" : I wonder what kind of files and what kind of contributions they are referring to ?

@mmonnin did the guy say anything about this in discord ?

And so if there are no files copied in that folder, what are the boinc tasks actually doing ?

In reply to JeromeC's message of 5 Feb 2026:

@Dave : Did you have a look at network bandwidth consumption ? are these tasks "streaming" anything ?

And regarding the "3. Place files in %USERPROFILE%\Axiom\contribute" : I wonder what kind of files and what kind of contributions they are referring to ?

@mmonnin did the guy say anything about this in discord ?

And so if there are no files copied in that folder, what are the boinc tasks actually doing ?

Yes he did mention that error. Since I believe the project is a waste of electricity I am not supporting uses on how to run it but its in discord a few days ago.

Minor Update .... my 38 tasks, from early on time wise, have moved from awaiting validation to valid with credit issue ...

Does not address lack of direct Admin communication or involvement... The version number of the project Applications has rapidly changed as the project has worked to address Application errors.

Bill F

And there's a cuda app now.

Just ran another batch. Not streaming anything. Tasks today however are using 100%CPU. I probably won't stick with the project though. Any new project I run in a VM till I see what it does with my system.

https://www.youtube.com/watch?v=rIW6iGqH0F8

https://www.wired.com/story/a-new-ai-math-ai-startup-just-cracked-4-previously-unsolved-problems/

It seems axiom is a startup focusing on mathematics, and they are apparently serious.

They just lack communication knowledge, at least for the boinc part.

It's a math term. I'm not sure that company and the project are related.

The base URL of the project is a Crypto coin exchange

https://heliex.net/

The GPU app was giving out some mega credit last night but it has been reducing over night. glibc 2.36 required (Ubuntu 24).

In reply to JeromeC's message of 7 Feb 2026:

https://www.youtube.com/watch?v=rIW6iGqH0F8

https://www.wired.com/story/a-new-ai-math-ai-startup-just-cracked-4-previously-unsolved-problems/

It seems axiom is a startup focusing on mathematics, and they are apparently serious.

They just lack communication knowledge, at least for the boinc part.

Very different projects.
That one is using AI to solve math problems.

This one is using BOINC to do AI training.


From their Project, About link.

What is Axiom?
Axiom is a distributed AI training platform that uses Hebbian learning — 'neurons that fire together, wire together' — to train a massive neural network across thousands of volunteer computers worldwide.

How It Works
Unlike traditional backpropagation, Hebbian learning is local and biologically inspired. Each sample strengthens connections between neurons that activate together — no complex gradient chains required.

The 85/15 Split: 85% of learning happens locally on your machine, while 15% comes from gossip synchronization with peers training other experts.

Getting Started
Download the Streaming Client — Get the Axiom Streaming Client (requires Python 3.8+)
Create an Account — Sign up to track your contributions and earn credits
Run the Client — Extract and run axiom_streaming_ui.py
Start Training — Click Start and watch your computer contribute to distributed AI!

Why Distributed Learning?
Modern AI training is concentrated in massive data centers controlled by a few corporations. Axiom explores an alternative: spreading the work across thousands of everyday computers, making AI research more accessible and democratized.

Your CPU performs real computations that contribute to training an open-source AI model. You earn credits for each weight synchronization, reflecting your contribution to the collective intelligence.

What Does My Computer Do?
The streaming client reads training data from your ~/Axiom/contribute folder (text files, documents, etc.) and performs Hebbian weight updates locally. Every 1000 samples, it syncs with the coordinator server to share learned patterns with other experts in the network.

The system monitors CPU and memory usage, automatically managing resources based on your preferences. An 'endocrine system' modulates resource usage — just like cortisol regulates your body's energy.

Contact
Questions or feedback? Post on the message boards or visit our GitHub.

All done in colours that make it impossible to actually read.

---

Grant
Darwin NT.

Thanks for making it readable.

And there is also this :

https://axiom.co/
https://axiom.co/customers/salad

That could have any relation or not relation.

It seems "axiom" it's a very fashionable word these days.