TLDR: Key Qubic Updates From the September 2026 All-Hands

• Neuraxon, Qubic's brain-like AI, reached Level 7 for the first time in the ARC-AGI-3 game lp85 on the day of the call, running entirely on standard processors.

• A fourth research paper was accepted at the Artificial Evolution conference (EA 2026) in Nice, France, held 27 to 29 October 2026.

• AEON Pay went live as Qubic's second ecosystem partnership, letting holders spend Qubic directly from their wallet.

• The MiCA entity is founded and settled, clearing the path toward listings on major European exchanges.

• Marketing moved fully to organic growth and held roughly 100K followers on X through August with no paid spend.

• The Qubic Q4 roadmap goes to a community vote on September 9, carrying the Neuraxon and ARC-AGI-3 research into the next quarter.

What the Qubic All-Hands Covered

Three groups presented at the September 3 Qubic All-Hands: the scientific team, marketing, and business development. Core Tech was not on the agenda this round, though its direction surfaced during the Q&A. What follows is a read of what moved, why it counts, and where each team is pointed next. If you are new to the project, the earlier All-Hands recap from August 6 gives useful background.

Scientific Team: Neuraxon, ARC-AGI-3, and Qubic's AGI Research

David Vivancos and Dr. Jose Sanchez opened with the update the community had been waiting on.

How Neuraxon Learns: Growing AI Instead of Training It

Neuraxon is the brain-like AI at the center of Qubic's AGI research. It is not trained the way a typical model is. It is grown. Many small versions compete, the strongest pass their traits forward, and the population improves generation after generation. This bio-inspired approach is what sets Qubic apart from language-model projects.

For the past two months Neuraxon has been learning to play ARC-AGI-3, an interactive benchmark widely regarded as one of the hardest tests in AI today. Rather than answering trivia, an agent has to explore an unfamiliar game and figure out the rules as it plays.

Neuraxon's ARC-AGI-3 Results at a Glance

The moment that drew the loudest response arrived live. After previously topping out at Level 4, Neuraxon reached Level 7 in one game for the first time during the session.

One point deserves underlining for anyone new to the project. None of this used high-end GPUs or large language models. Neuraxon ran on ordinary processors and still outscored Grok 4.2 and several GPT versions on the games it played. That keeps costs low, and it shows Qubic's grow-it approach can compete on its own terms. The competition's final games open on November 2, and the team plans to publish the full data and code once the contest closes. For a plain-language look at how Qubic thinks about measuring intelligence, the Neuraxon Intelligence Academy piece on the g factor and ARC-AGI is a good primer.

Peer-Reviewed Research and the Neuraxon Game of Life

The fourth accepted paper heads to EA 2026 in Nice this October, one of 22 selected for oral presentation. It follows the team's recent showing at AGI-26 in San Francisco, where they presented the Multi-Neuraxon model. Both milestones are covered in more depth in Qubic's Open Science AMA recap.

Anyone can watch the research in motion through Neuraxon GoL Live, a Game of Life simulation that runs around the clock. Hundreds of iterations and thousands of Neuraxons live, compete, and pass their traits to the next generation on a public leaderboard. Each cycle feeds the upcoming Neuraxon 3.0.

Qubic's Q4 Roadmap: Neuraxon 3.0, Predictive Coding, and Anthill

The Qubic Q4 roadmap keeps ARC-AGI-3 at the center and adds Neuraxon 3.0, a build that gathers what the team has learned into a single model to test against the benchmark and inside Anthill once it reaches the network. Predictive coding is the next idea in focus, the brain's habit of anticipating rather than reacting. A quick example makes it concrete: when you reach for a bottle, your brain predicts its weight before your hand arrives. That kind of prediction is what the team wants its models to learn.

Marketing: Organic Growth and Community Engagement

Marketing walked through a clean reset. Paid media spend went to zero, and the strategy moved to organic growth. The numbers backed the decision. Through August, with nothing spent on ads, the following on X held near 100K as inactive bot accounts fell away. Impressions reached 452.2K, engagement climbed across the board, and on August 26 Qubic ranked as the second most-posted asset on X. The community reacting to Qubic is real, and it is being reached at a small fraction of the earlier cost.

The larger shift is in approach. Rather than promoting Qubic in the abstract, marketing now works alongside each team to tell the specific story of what they build. Early examples include backing business development's AEON campaign and reshaping the science team's NIA Academy into structured learning content. A dedicated science podcast is also in progress, giving the deep research its own space and an audience suited to it. Community Spaces, run with the community team, are slated for this month. On the growth side, Cointribune is collaborating with the team to lead SEO and write articles for the website, and marketing is handling website improvements. The organic footing is expected to hold at least through the year-end.

Ecosystem and Business Development: AEON Pay, MiCA, and Mining

The treasurer also delivered the business-development update, and adoption was the theme.

AEON Pay is Qubic's second ecosystem partnership, building on the earlier payments work covered in the August All-Hands recap. It handles the payment card and checkout while holders spend Qubic straight from their wallet, so paying no longer requires a manual step into dollars or a stablecoin first. Wallet to checkout, in one flow.

On regulation, the entity that will carry Qubic through Europe's Markets in Crypto-Assets (MiCA) framework is founded and settled. With that structure in place, the team can now file the paperwork that clears major European exchanges to list the asset. Developer activity is being tracked across three groups, and a regulatory unblock is progressing, with fuller detail to follow directly from Kim.

Following Apool's move away from Qubic mining, recruiting new mining pools and miners became a clear near-term priority alongside the MiCA work. GridHash and the wider team are aligned on it, and marketing is already lined up to help bring new participants on board. Mining sits at the heart of the network because Qubic runs on Useful Proof of Work, which points mining energy toward AI training.

Core Tech and Aigarth: Where Qubic's Roadmap Is Heading

Core Tech was not on this agenda, and its direction came through during the Q&A. Training stays in-house for now, which keeps the research focused while the team builds toward Aigarth, the layer designed to let anyone grow their own AI on Qubic. Questions about an outside AI kit and future commercial models point to where the roadmap is heading, with product details taking shape as the technical foundation matures. The tone matched the wider project: a research effort moving with intent, each piece coordinated rather than rushed.

Q&A Highlights: Neuraxon vs. LLMs

The clearest explainer of the day answered a simple question. What makes Neuraxon different from a large language model? The team reached for a metaphor. To fly, you can engineer a plane or you can evolve a bird. Language models are the plane, built from language and probability across enormous training sets. Neuraxon is the bird, grown from biology, where prediction comes first and there is no training phase in the usual sense. Both are impressive. Qubic simply finds the biological path the more natural one to explore.

A second question asked whether the ant-colony work and Neuraxon might eventually strengthen each other. That is an open research question the team is now set up to explore, and the first step is scale. Today's experiments use models as small as 150 neurons, and strong results at that size are a promising sign. Growing them toward a few thousand neurons is where the team expects new structures to appear and accuracy to improve. The July 9 All-Hands recap traces how these benchmark results have built over the summer.

Looking Ahead: Qubic's Next Quarter

September 3 showed a project advancing on several fronts at once. The science team has a benchmark that keeps climbing and a fourth peer-reviewed paper heading to Nice, with its living simulation online for anyone to watch. The ecosystem has a working payment rail and a settled path through European regulation. Marketing has proof that the community is real and a sharper way to serve it. The vote on September 9 hands the next quarter to that community, and the work leading into it is already funded and moving.