Optimizing Contribution at Scale: Fairer Job Assignments for Provers

We’ve made prover job assignments more fair, stable, and resistant to abuse, rewarding real contributors and improving the health of the network.

As the Nexus community continues to grow, we are always looking for ways to make it easier and more rewarding to contribute compute.

Today we have improved how prover jobs are assigned, moving from a quantity-based race to a reputation-driven model. This ensures reliable contributors receive steadier, higher-value work while the network stays resilient under heavy load. Below, we cover what is changing, why it matters, and how to make the most of it.

What’s changing

We have introduced a reputation-based job assignment system that prioritizes quality over quantity and smooths network performance during peak demand.

• Reputation-based throttling: Jobs are now assigned based on the reputation of each session, not just the user or number of nodes.
• Per-session scoring: Good behavior (valid proofs, steady uptime) raises your reputation score and unlocks larger, higher-value tasks, while poor behavior lowers it.
• Task normalization: Jobs are bucketed so similar machines get comparable workloads.
• Concurrency limits: Prevents any single machine from monopolizing the queue.
• Abuse detection: Repeated invalid proofs or churny connection patterns reduce priority.

Reputation details: Every proving session starts at a baseline reputation. As you complete valid work, your reputation will increase, unlocking larger jobs.

For CLI v0.10.4 and later:

Older CLI versions cap at 10x jobs even with great reputation. Only v0.10.4+ can request the largest tasks. Reputation is session-based for now, so restarting resets the score. Task size will still vary, but higher reputation steadily increases access to high-value work.

A Guide to the Nexus CLI

The Nexus Command-Line Interface (CLI) is the fastest on-ramp for contributing compute to the world supercomputer.

NexusDaniel McGlynn

Why this matters

The old approach rewarded node quantity over quality. Users could spin up hundreds of low-power nodes and still receive as many jobs as high-quality contributors. This led to congestion, instability, and slower assignments for everyone.

The new model delivers:

• Fairness: One session, one reputation. Spamming nodes no longer offers an advantage.
• Reliability: Predictable assignments replace the fastest grab race.
• Efficiency: Better distribution reduces errors and wasted effort.
• Better rewards for real work: Stable setups earn fewer but higher-value tasks, giving genuine contributors a path to better returns.

What you can do right now

These steps will help you benefit from the new scheduler immediately.

Run one solid Nexus OS node or a CLI node with at least 4 GB of RAM, and keep it running steadily to build your session’s reputation. Avoid requesting more work than you can complete, and update to CLI v0.10.4+ to unlock the largest possible rewards.

Best practices to maximize points

Building a high-reputation track record is the best way to consistently earn high-value jobs.

Each CLI process has its own session reputation. Focus on maintaining a few high-reputation instances rather than spinning up many new ones at zero reputation. Give each CLI enough memory and CPU so it can complete every task without errors.

One stable, high-reputation node earning large tasks is better than many low-reputation nodes struggling to keep up. Remember that a lower-point task does not always mean your reputation has dropped, as task size and length vary naturally.

Looking ahead

This new system lays the groundwork for a healthier, more scalable network. It means higher-value jobs for proven contributors, a transparent and resilient reputation model, and stronger protections for every user.

We are moving toward a future where real work is rewarded, spam is neutralized, and every contributor gets a fair shot.