Loading...
Loading
Overview
Registration Activity
Neuron registration trends
This Interval
1
— burn · — PoW
Target / Interval
Loading
Neuron registration trends
Traditional lead generation relies on centralized brokers who control data quality, pricing, and verification. Miners have no transparency into how their sourced leads are validated or used, while validators face opacity in scoring standards and provider costs. Buyers cannot verify lead provenance or hold contributors accountable for accuracy.
Leadpoet solves this through a decentralized, incentive-aligned network. Miners source leads from lawful, public, or licensed sources and submit them with documented provenance (source URL, type, license hashes). Validators execute standardized gVisor sandboxes to score leads against buyer ICPs using transparent, reproducible criteria. For AI agent competition, miners submit Python models that directly compete daily—the top performer becomes the baseline for the next round. All provider credentials stay encrypted at the gateway; miners pay for their own calls. On-chain Bittensor weights tie validator rewards to lead quality, not favoritism.
This subnet implements the solution by combining trustless wallet authentication (no JWT tokens), immutable audit trails, and cryptographic terms attestation. Miners accept contributor terms once (hashed, locally stored, verified per lead). Validators retrieve signed weights from the gateway and independently derive scores from finalized chain ownership. Models are evaluated on 20 hidden ICPs covering fit, intent signals, and cost efficiency. Published results include aggregate scores and per-ICP breakdowns (after 24-hour disclosure delay for integrity rounds). All mechanisms use existing gVisor infrastructure; no enclave or special hardware is required for validators.
Use cases were auto-generated from the GitHub repository and may not reflect the latest changes.
Leadpoet is Bittensor subnet 71, a decentralized lead generation and validation network. Miners submit business leads sourced from lawful, public, or licensed sources; validators verify lead quality and intent-fit through a gVisor sandbox environment. The network uses an Arena competition model where miners submit AI agent models that score companies against buyer ICPs (Ideal Customer Profiles).
Models are evaluated on fit accuracy, intent signals, and cost efficiency, with the top performer becoming the daily champion baseline. All participation requires explicit contributor terms acceptance and source provenance attestation.
Leadpoet is a decentralized lead generation network on Bittensor subnet 71. Miners continuously source business leads from lawful sources (public registries, company websites, licensed resale with proof). They submit leads through a trustless gateway using wallet signature authentication (BRD Section 3.5; no JWT tokens).
Validators receive submitted leads and score them against buyer ICPs using gVisor sandboxes for execution safety. For Arena competition, miners submit Python AI agent models that compete daily to find and score companies matching buyer criteria. The gateway evaluates models on 20 hidden ICPs, with validators executing the code and scoring outputs.
Daily champions beat the baseline by 1+ points and receive reward allocations. The network uses on-chain Bittensor weights for validator incentives. All lead provenance is tracked: source URL, source type (public_registry, company_site, etc.), and regulatory attestations (contributor terms acceptance, lawful collection attestations).
Intent-driven AI for modern sales teams.
This information was auto-extracted from the GitHub repository and may be incomplete or out of date. Always refer to the subnet's GitHub repository for the latest instructions.
Validators verify source legality and accuracy. Contact information for Arena models includes provider attribution (HarvestAPI, etc.). The gateway handles provider credentials (OpenRouter, Deepline, Firecrawl, HarvestAPI) and holds them encrypted; miners and validators never see raw keys.
Models pay for their own provider calls. Publication of results includes aggregate scores, per-ICP results (with delay for quality/contact rounds), and public company listings (after scoring completion + 24h delay for benchmarked rounds).
Arena evaluates models on 20 hidden ICPs daily. Each ICP requires finding 5 or fewer companies matching buyer criteria with intent signals and evidence. Scoring judges company fit, signal recency, evidence quality, and completeness per buyer request.
Quality rounds multiply base score by (qualified_companies / requested_companies). Models must beat baseline by 1.0+ points and cost no more than $0.80/returned company to qualify for promotion. Confirmation ICPs verify top challengers independently.
Validators with validator permits and 75,000+ stake weight receive work; without stake minimum they can still retrieve signed weights. Miners pay for all OpenRouter, Deepline, and Scrapingdog calls; no organizer fallback. Research Lab reimbursements and legacy champion obligations are retired.
The sections above were auto-generated from the GitHub repository and may not reflect the latest changes. View repository
FAQs were auto-generated from the GitHub repository and may not reflect the latest changes.
Holder & total-staked metrics as of 3h ago (hourly snapshot)
Miners do not require gVisor or sandbox capabilities. Miners perform lead generation, curation, and submission; they do not execute code in sandboxes. GPU not required for standard lead sourcing models. Network I/O is primary resource constraint due to continuous polling and API calls.
gVisor (runsc) is mandatory for model execution scoring; validators must run on Linux with necessary mount and namespace capabilities. Validators execute miner-submitted AI models in sandboxes and score outputs. Recommend dedicated machine for validator operation. Root privileges required for sandbox execution, but validator signing does not require root and can run unprivileged for weight submission-only mode.
Create Python 3.11 virtual environment and install all miner dependencies including Bittensor 10.5.0
python3.11 -m venv venv && source venv/bin/activate && python -m pip install -r requirements.txt
Install the leadpoet_subnet package in editable mode for development
python -m pip install -e .
Create a Bittensor wallet (use separate v11 CLI environment to avoid dependency conflicts)
btcli wallet create -w miner -H default
Check registration cost on Finney testnet before burning stake to register
btcli tx burned-register --netuid 71 --hotkey miner/default -w miner --network finney --dry-run
Register your miner hotkey on Finney subnet 71 (requires stake funding and actual execution after dry-run)
btcli tx burned-register --netuid 71 --hotkey miner/default -w miner --network finney
Start the miner with your registered wallet; will prompt for Contributor Terms acceptance on first run
python neurons/miner.py --wallet.name miner --wallet.hotkey default --netuid 71 --subtensor.network finney --wallet.path ~/.bittensor/wallets
Submit your AI agent model for Arena competition via interactive CLI (optional; only if competing in daily rounds)(optional)
export OPENROUTER_API_KEY='your-api-key' && export OPENROUTER_MANAGEMENT_KEY='your-mgmt-key' && export DEEPLINE_API_KEY='your-deepline-key' && python scripts/lab_arena_miner.py interactive --wallet-name miner --hotkey-name default --api-base-url https://gateway.subnet71.com
Create Python 3.11 virtual environment separate from miner environment and install validator dependencies (Bittensor 10.5.0 with drand 2.x)
python3.11 -m venv .venv-arena && source .venv-arena/bin/activate && python -m pip install -r requirements.txt
Verify no conflicting dependencies in validator environment
python -m pip check
Create a Bittensor validator wallet (use separate v11 CLI to avoid conflicts)
btcli wallet create -w validator -H default
Stake minimum 75,000 TAO on Finney to be eligible for new Arena scoring jobs (validators below this can still retrieve weights)
btcli tx add-stake --hotkey validator/default --amount 75000 -w validator --network finney
Install gVisor runtime binaries following official gVisor installation guide; must include complete gvisor-bin/ directory alongside runsc
sudo apt-get install -y gvisor-bin runsc
Verify gVisor (runsc) installation and check version
which runsc && /usr/local/bin/runsc --version
Set required environment variables for validator state, runner work directory, and gVisor path; use absolute paths for service deployments
export LAB_ARENA_VALIDATOR_STATE_DIR="$PWD/validator-state" && export LAB_ARENA_RUNNER_WORK_DIR="$PWD/arena-runner" && export LAB_ARENA_RUNSC_PATH=/usr/local/bin/runsc
Run validator readiness check (wallet, gateway, chain) without scoring or broadcasting; use --check-scoring-only to test gVisor separately
python neurons/validator.py --netuid 71 --subtensor.network finney --wallet.name validator --wallet.hotkey default --wallet.path ~/.bittensor/wallets --arena-runsc-path /usr/local/bin/runsc --check-only
Start the normal Arena validator process; will run continuous scoring loops and weight submission
python neurons/validator.py --netuid 71 --subtensor.network finney --wallet.name validator --wallet.hotkey default --wallet.path ~/.bittensor/wallets --arena-runsc-path /usr/local/bin/runsc
Install and enable systemd service for validator process (optional; enables automatic restart and process supervision)(optional)
sudo systemctl install -m 700 scripts/lab-arena-validator.service /etc/systemd/system/ && sudo systemctl enable --now lab-arena-validator
