MMoletariat

What we can say today

Intelligence that learns its way forward.

The work is still young, but Arina has already demonstrated a set of capabilities that are easier to understand in human terms than in research jargon.

  1. Learns

    Builds a picture of the world.

    Arina can turn experience into a working model of what is happening, then revise that model when the evidence changes.
  2. Reasons

    Knows when the answer is not settled.

    She can preserve uncertainty, compare competing explanations, and seek information that would make a real difference.
  3. Plans

    Thinks beyond the next move.

    Arina can consider possible futures, weigh their consequences, and plan when the world is only partly understood.
  4. Discovers

    Invents and reuses useful ideas.

    She can form new concepts and methods, test whether they work, and carry what she has learned into unfamiliar problems.
  5. Explores

    Makes mathematical discoveries.

    In controlled research settings, Arina has rediscovered mathematical distinctions, proof strategies, and useful constructions.
  6. Communicates

    Can learn shared language from interaction.

    She has learned grounded signals and simple grammar by working with another agent, without beginning from human language.
  7. Reflects

    Can examine her own failures.

    Arina can identify gaps in her abilities, propose improvements, and test whether a repair genuinely solves the problem.
  8. Improves

    Keeps self-improvement under human control.

    Arina can diagnose and test changes to its own machinery. Recursive use is available for controlled testing, while promotion remains behind external oversight.

Why Moletariat

Capability should come with public purpose.

If general intelligence changes the shape of work, discovery, and opportunity, its gains should not belong only to the organisations wealthy enough to build it. Moletariat exists to pursue powerful AI with accountability, broad benefit, and a place for ordinary people in the future it creates.

Where things stand

Strong evidence. Deliberate limits.

Arina is moving quickly, but the claims should move only as fast as the evidence. These are the boundaries around what we can responsibly say today.

  • Arina is a research programme, not a publicly available AGI product today.
  • We are not claiming consciousness or human equivalence.
  • We have deliberately chosen not to run recursive self-improvement in the live system. The machinery is available for controlled testing, but remains behind an external governance boundary.
  • Current results come from controlled research environments; wider real-world capability still has to be earned.

That said

The pace is real.

Progress has been strong, and so has the evidentiary output: claimed steps are accompanied by tests, preserved failures, and reproducible results.

With more resources behind the programme, we want to push further into the areas where controlled research can begin meeting the wider world.

  • Real-world perception
  • Open mathematical discovery
  • Natural language
  • Scientific discovery
  • Independent evaluation
  • Governed self-improvement
Help shape what comes next

Contact

Get in touch.

For research, collaboration, press, or general enquiries about Arina and Moletariat, write to us directly.

hello@moletariat.ai