AXONRY / WEIGHT-SPACE PROGRAMMING

Intelligence,
rewritten.

Read the weights. Rewrite the behavior.
Explore direct model editing on Bittensor.

Explore weight space

01 / READ

Read the weights.
Find the change.

Give a model an objective. Let it identify the structures in a policy network that shape behavior.

BASE MODELW₀Weight structure

02 / WRITE

A small patch.
A new policy.

Turn an objective into a local weight update. A small parameter change can shape a new policy.

W₁ = W₀ + ΔW

03 / EXECUTE

Same task.
New behavior.

Apply the patch and run the obstacle task again. Watch a local weight update become a new behavior.

04 / VERIFY

Every edit.
Proven again.

Axonry rebuilds the edited model and measures it against the current champion.

  • Benchmark performance
  • Capability regression
Explore the subnet

AXONRY / BITTENSOR SN99

A new paradigm for decentralized LLM development.

Our vision is decentralized LLM training. Today, miners propose ways to improve a shared language model through direct weight edits, without gradient updates.

  • ARBOS

    A framework for frontier agents

    Arbos runs agents in parallel. Each agent tests a miner’s method in a dedicated GPU sandbox and returns a weight edit.

  • AXONRY PROTOCOL

    A protocol for model improvement

    Axonry brings Bittensor miners together under shared rules, schedules their experiments, and benchmarks each edit against the current champion.

The long-term goal is thousands of isolated experiments running in parallel.

Weight-space research

7.9% → 48.1%MATH-500 accuracy

Editing a small fraction of Qwen3.5-0.8B weights reduced reasoning loops in the original research experiment.

4 / 4Game policies identified

Astra inferred the objectives of four policies from their weights and input and action definitions.

3 / 3Games above the PPO baseline

Three policies written directly by Astra outperformed the trained baseline in MinAtar.