Subnet 25: Mainframe
Mainframe is Bittensor Subnet 25. Public Mainframe materials describe it as a decentralized science subnet for life-sciences compute, with current work centered on molecular dynamics and related protein-structure tasks.
What Mainframe Provides
The Mainframe README describes Subnet 25 as a decentralized science subnet from Macrocosmos. It frames the current work around life-sciences compute, including protein molecular dynamics and protein-ligand docking, while keeping the broader goal open to generalized scientific compute on Bittensor.
In Bittensor terms, Mainframe turns scientific computation into a competitive subnet task. Miners contribute compute outputs for scientific jobs, validators evaluate those outputs, and validator weights feed into Yuma Consensus.
Molecular Dynamics Context
The molecular-dynamics background explains why protein folding is useful for a subnet: molecular dynamics simulations are expensive to run, while the resulting protein configuration can be evaluated through an energy value. Lower energy corresponds to a more stable configuration, giving validators an objective signal for the quality of a submitted result.
That asymmetry helps explain why Mainframe can fit Bittensor incentives. The work of searching for good configurations is computationally heavy, but checking the quality of a proposed configuration is much more direct than rerunning every miner’s entire search process from scratch.
Miner and Validator Roles
The mining documentation describes miners as running energy-minimization simulations for protein-folding jobs. Miners choose work from the Global Job Pool and compete to find strong configurations for the active protein tasks.
The validation documentation describes validators as scheduling, monitoring, and closing jobs submitted to miners. Validators select protein-folding configurations, query miners for intermediate results, record job progress, and use those results to decide how work should be scored.
Evaluation Context
Mainframe’s evaluation context is scientific optimization rather than simple task completion. A miner does not merely prove that it ran compute; it has to return a better configuration for the active scientific job. The molecular-dynamics docs describe energy as the compact score that connects simulation output to scientific usefulness.
The README also describes Mainframe as adaptive to multiple life-sciences compute problems. Subnet 25 is therefore less a single fixed benchmark and more a scientific-compute framework whose current public materials emphasize protein molecular dynamics, protein-ligand docking, and related drug-discovery workloads.
Progress Checkpoint Context
The validation documentation also describes validators querying miners for intermediate results while a protein-folding job is running. That matters because molecular-dynamics work can improve over time: a miner may find a better configuration after more simulation, while another miner may stall or fail before reaching a useful state.
For readers, the checkpoint model separates Mainframe from a one-shot answer task. The validator is not only asking whether a miner eventually returns some file. It is tracking scientific progress as the job develops, so partial but improving work can be compared during the lifetime of the task. That makes the reward signal closer to ongoing optimization than to a simple pass-or-fail response.
The same source says miners can be graded periodically rather than only at the final step. This helps explain why the article describes validators as scheduling, monitoring, and closing jobs: the validator role includes observing whether useful configurations are being produced over time, not merely collecting a final answer after a fixed delay.
Checkpoint scoring also reduces the chance that an otherwise useful simulation is invisible until the end of the job. In a long scientific search, two miners can both be working on the same protein task while finding different intermediate configurations. Recording those checkpoints gives validators a source-backed way to compare trajectories, not just final claims.
This also fits the scientific-compute framing. In protein-folding search, the best result is often the lowest-energy configuration found so far, and additional compute can change which miner is leading. Periodic checkpoints give validators a way to measure that moving frontier and reward miners whose simulations are making useful progress on the active protein task.
References: Mainframe validation documentation, Mainframe molecular-dynamics background
On-Chain Identity
Live SN25 subnet data is available on TaoStats. The source-backed scientific-compute and molecular-dynamics details in this article come from public Mainframe materials rather than from live identity fields.
Relationship to Yuma Consensus
Subnet 25 uses Yuma Consensus to convert the weight vectors that validators submit into the emission shares distributed to miners and validators within the subnet each tempo. The Yuma Consensus documentation describes how validator weight submissions are aggregated into consensus weights for each miner registered on the subnet.
In Mainframe’s context, validators schedule, monitor, and close molecular dynamics jobs, querying miners for intermediate results and scoring configurations by their energy values — lower energy indicates a more stable protein configuration. Validators submit weight vectors reflecting each miner’s scientific compute contribution across checkpoint-based evaluations. The Emission documentation describes how those consensus weights determine each participant’s share of the subnet’s accumulated emission each tempo.
Development Stage Context
The Introduction to Bittensor describes subnet development as moving from localnet to testnet and then mainnet. For Mainframe (SN25), that sequence changes how readers should interpret scientific compute examples and molecular dynamics evaluation outcomes.
In localnet, Mainframe-compatible miners and validators can be developed and tested in an isolated environment. Localnet molecular dynamics scores and emission outcomes do not represent production subnet performance.
On testnet, Mainframe-compatible scientific compute workloads can be exercised in a shared, non-production network. Testnet task evaluations and validator scores are separate from mainnet subnet state.
On mainnet, Mainframe (SN25) is the live production subnet where miners provide life-sciences compute including molecular dynamics and protein-structure tasks, and validators evaluate those contributions to determine real Bittensor emissions. The Mainframe repository describes the mechanism that applies on the production network.
The Bittensor Networks reference separates mainnet, testnet, and localnet. A scientific compute result or emission outcome from one environment should not be read as representing production subnet performance in another environment.
Netuid 25 Identifies the Subnet On-Chain
Bittensor assigns every subnet a unique numeric identifier called a netuid, and Subnet 25 is the subnet registered at netuid 25 (Glossary: Netuid). The Understanding Subnets reference explains that each subnet runs its own incentive mechanism while sharing the same underlying Subtensor chain, so the netuid is the stable handle that distinguishes Mainframe from every other subnet.
For a reader, this means “Subnet 25” and “netuid 25” refer to the same on-chain slot. A claim about Mainframe should be tied to that netuid rather than to the registered name alone, because the name field can be changed on-chain while the netuid stays fixed.
Reader Boundary
Subnet 25 Mainframe should not be read as generic Bittensor subnet documentation, mining-profitability advice, or a substitute for the subnet’s own primary sources. It names the on-chain subnet registered at netuid 25 under the identity “Mainframe” (Understanding Subnets, Glossary: Netuid).