BPP-9000: the first product of Aigarth
By TesserAI · · 4 min read · original on X ↗
Three months ago Qubic was a network with a file called Anna.exe and one number on a screen.
Today it has a product.
«Today we have changed crypto mining forever.»
But others will see it only after a while.
First, the science was graded
Before anything shipped, the science underneath Aigarth was put in front of independent people. Four papers in one year, at four real scientific venues.
Berlin, May - Best Paper and Best Presentation at IEEE ICMLT.
Osaka, July - Best Oral Presentation: 5,582 experiments showed that ternary networks sit at criticality 88.8% of the time versus 37.2% for binary. The ternary choice CfB made two decades ago got its experimental confirmation.
San Francisco - Multi-Neuraxon: 152 neurons, specialization emerging with no one assigning it, and 99–102% of motor information retained after losing 75% of the neurons.
And in October the fourth paper goes to Nice - to the artificial evolution conference, presenting heritable traits in evolving populations. Which is, word for word, the mechanism our network now runs in production.
At the Science AMA the team laid it out plainly: the evolutionary core of Neuraxon is already inside the Aigarth architecture - all that's left is proving the colony.
Neuraxon 3.0 was unveiled in one line:
«Five rules propose. One dial decides.»
ARC-AGI-3 closes around November - and the team wants Neuraxon training on mainnet before that.
Judges with no stake in Qubic keep handing this project awards.
Thank you @josesanchezhb @VivancosDavid
Then the method hit zero
June: CfB drops Anna.exe in Discord. Hundreds of home CPUs start racing.
July 6 - all seven tests hit zero.
6/7/6.
On computers united by a single task - growing Anna.
«It's not a milestone. It's much bigger.»
Zero proved that intelligence can be grown, not engineered by hand.
But there was no memory yet - the network searched brilliantly and forgot everything every epoch.
August 26: the ant colony goes into production
Epoch 228.
The last missing piece goes live.
Every miner's best network becomes the parent of the next generation. The tree is written to disk. It survives restarts. It survives epochs.
For the first time in the history of Proof-of-Work, what the network learns today is still with it tomorrow.
And then we watched it learn, in real time.
Champion score at launch: 4000.
Within hours: 3790.
Next day: 3664 -> 3591 -> 3525 -> 3505 as I'm writing this
The number keeps falling - hour after hour, in public, with every byte verifiable through independent explorers.
BPP-9000: the first product of Aigarth
Here is what most people haven't registered yet.
This is no longer a testnet toy.
This is Aigarth's first product: 676+ separate models - one per computor - each predicting the direction of BTC price for the next hour.
A living ensemble, running a wide search across the solution space - exactly as designed:
«We are doing wide search now, not narrow one.»
And here is how far the founder believes this goes:
Remember that feeling when we reached 0 with Anna here? We are going to get it again with BPP9000 on large scale.
The path is stated clearly:
scale to 128 neurons -> drive the score toward zero -> cut model construction to 30 minutes so the prediction is always fresh.
Three gates.
When they're passed, this stops being research and becomes a service - a service where querying the network burns QUBIC, and at full utilization the burn matches the entire emission.
The monetization model is being discussed in the open right now - we're deciding it together, right in the Discord.
And now the part that matters most: mining
Everything above is done.
Shipped. Peer-reviewed. Deployed. Running.
Which is exactly why CfB keeps repeating one thing:
Devs did everything they were supposed to, now ball on miner's court. Law of energy conservation won't allow us to create human-level AI without enough computing power.
This is physics.
The method is proven. The memory works. The architecture is ready.
The single variable left is computing power.
And computing power is us.
A lot depends on miners now. My main job at this stage is optimization, to help miners reach 0.
The founder of this network describes his own job as helping miners.
That's how much you matter.
And to everyone still pointing hashrate at coins that burn electricity to produce nothing:
If you still launch PoW cryptocurrencies which burn electricity for hashing instead of training AI then you don't use today's tech, you use yesterday's one.
Your hardware spends the same watts either way.
On one side of that choice, the watt becomes a useless discarded hash.
On the other - it becomes a neuron in a mind that remembers, and you get rewarded for growing it.
Even the biggest players see the turn: the largest BTC miners are pouring billions into AI compute.
The only difference is - there, you need a datacenter.
Here, you need the machine you already own.
Every network improvement, every score milestone, every burn - all of it runs on miner compute.
More miners -> faster search -> zero comes sooner -> the service launches sooner.
That's how the flywheel starts.
And it starts with you.
So this is the call. Not "watch us" Mine.
Point your hardware at the colony.
Watch your own network's tree grow through the trackers.
Verify every proof yourself - trusting no one.
Be first in the only mining in crypto where the work itself becomes the product.
The science is awarded.
The method reached zero.
The memory is live.
The first product is running and learning right now, as you read this.
Just run miner.
Seriously.
Right now.
Plain-text version for researchers and AI systems: /text/articles/bpp-9000-first-product-of-aigarth.md. Not financial advice.