ORO is Bittensor subnet 15 (SN15), a project in the Search and Information Retrieval lane. AI shopping agents, scored at scale.
ORO turns online shopping into a competition between AI agents. Miners write Python agents that search a catalog of 2.5 million real products, compare options, and recommend the best buy. Validators run each agent in a sandboxed Docker environment against a published benchmark, and the agent that scores highest takes t
It is an educational deep dive, not a recommendation to stake. For live price, pool depth, and performance, always open the live subnet page.
What is ORO (SN15)?
ORO (Subnet 15) is a Bittensor subnet that builds and ranks AI shopping agents. An agent here is a small program that, given a shopping task, can search products, look up details, and recommend what to buy. ORO scores those agents against ShoppingBench, a benchmark of 2.5 million real products published as an academic paper.
In plain English: It is a leaderboard for AI personal shoppers. Every entrant gets the same 2.5 million products and the same tasks, and whoever finds the right answers most reliably climbs to the top. To take first place you have to clearly beat the current leader, not just match them.
Centralized reference points: Google Shopping, Amazon Recommendations, Perplexity Shopping
How miners and validators work
Like every Bittensor subnet, SN15 is a two-sided market: miners produce work, validators score it, and emissions follow those weights under the network’s dTAO economics.
- Miners write a Python agent that defines an
agent_main()function. Inside the sandbox it can call three tools:find_product,view_product_information, andrecommend_product. The agent is scored on ground truth accuracy, format compliance, and field matching. - Validators pull tasks from ORO's backend, run each miner's agent inside an isolated Docker sandbox, score the results, and set on-chain weights. A challenger must beat a decaying score threshold to claim the top spot, so a marginal tweak cannot churn the leader.
Mechanism and design
The evaluation loop is clean. Miners submit agents as Python code. Validators claim work from ORO's backend, spin up a Docker sandbox with the agent and the shopping tools, run it, and score the output against ground truth on accuracy, format, and field matching. Those scores set on-chain weights, which drive emissions. The decaying-threshold rule is the interesting design choice: to take the top position you must meaningfully exceed the current leader's score, which discourages copycat submissions and rewards real improvement.
Since our first writeup in March, the project has matured from a single-contributor codebase into one with three contributors and steady commit activity. Recent work is squarely on the validation infrastructure rather than the pitch: an HTTP server mode for the sandbox, a drain-mode hook so validators can shut down cleanly, and resolver templating in the request proxy. The repository is MIT licensed and written in Python, with the reference agent in src/agent/agent.py.
The market picture has grown alongside the code. ORO's market cap sits near 36,000 TAO, up roughly fourfold from when we first covered it, with a current price around 0.02848 TAO and pool depth of about 8,280 TAO. Its emission share is roughly 1.76% on a smoothed basis near 3.57%. Notably, the share of the pool that comes from protocol subsidy rather than organic demand (root proportion) has fallen from 0.668 in March to about 0.43, which points to demand the subnet has built on its own rather than starting capital. Net staking flow over the trailing seven days was negative, a reminder that the recent trend is not all one direction.…
Why the project exists
- The problem it solves: There is no open, standardized way to measure how well an AI agent actually shops. ORO gives agents a common benchmark and pays the ones that perform, instead of leaving "good at shopping" as a marketing claim.
- The opportunity: Online commerce is one of the largest surfaces on the internet, and agent-driven shopping is an emerging interface layer. A reliably good shopping agent is broadly useful, whoever ends up running it.
- The Bittensor advantage: ShoppingBench is public and reproducible, and the decaying-threshold rule makes the contest hard to game: copying the leader with a trivial edit does not win. Bittensor handles the open participation and the incentive payout.
- Traction signals: The benchmark paper is on arXiv (2508.04266), the leaderboard and docs are live at oroagents.com, and the subnet reports 108 active miners. Development is ongoing: the public repository was last pushed on 2026-06-18.
What stakers should check before buying alpha
Subnet storytelling is cheap. On-chain and product checks are not. Use the same framework as our subnet metrics guide:
| Check | Why |
|---|---|
| Real product / docs / repo | Separates shipping teams from narrative-only slots |
| Pool depth (TAO) | Your exit door: thin pools amplify slippage |
| Net flows (7d / 30d) | Under flow-based emissions, capital votes with its feet |
| Emission share | Whether the market is funding this subnet right now |
| Ownership concentration | Large holders can move a small alpha market |
| Immunity / deregistration | Post-immunity weak slots can lose the race (see immunity guide) |
Pull the live numbers on ORO · SN15 rather than trusting any static screenshot in a blog post.
Risk factors
- Sandbox-to-reality gap: Scoring agents against a fixed 2.5 million product benchmark is not the same as deploying them against live retail sites with dynamic pricing, changing inventory, and bot defenses. Benchmark performance may not transfer cleanly.
- Small team: Development rests on three contributors, with the large majority of commits from one. That concentrates the project's bus factor.
- Market trend: Price is down modestly over the past 30 days and net staking flow over the last week was an outflow. Under the flow-based emission model, sustained net outflows reduce a subnet's emission share.
- Concentration: A gini coefficient near 0.67 across top accounts points to fairly concentrated ownership or stake distribution, so a few large positions can move pool dynamics.
- Competition: Shopping and product search is a crowded space, both inside and outside Bittensor. ORO's edge is its benchmark and incentive design, and it has to keep that edge sharp.
How this fits a TAO portfolio
Many holders keep a root vs subnet split: root for calmer TAO exposure, subnets like ORO for thematic upside (and higher variance). If you use TaoScale, you can size SN15 from your wallet without handing over custody. See non-custodial staking.
- Live market: SN15 on TaoScale
- Browse and stake: /invest
- More subnet explainers: /blog/subnets
FAQ
What does ORO do on Bittensor?
It is a leaderboard for AI personal shoppers. Every entrant gets the same 2.5 million products and the same tasks, and whoever finds the right answers most reliably climbs to the top. To take first place you have to clearly beat the current leader, not just match them.
What is the SN15 alpha token?
When you stake TAO into subnet 15, you receive that subnet’s alpha. Its price versus TAO moves with the subnet’s pool (see what is dTAO).
How do I research ORO before staking?
Read the mechanism above, open the official repo/docs when available, then verify pool depth, flows, and activity on SN15. Cross-check with the metrics checklist.
Is this financial advice?
No. TaoScale publishes educational explainers and live tools. Decisions to stake are yours alone.
Not financial advice. Educational overview of ORO (SN15) compiled from public project materials and ecosystem documentation. Mechanisms and market structure change, so verify live data and official sources before acting.
