Most people think “oracle = Chainlink”. But there’s a big architectural difference between Chainlink’s curated DONs and a potential Kaspa-style PoW-native oracle where miners themselves attest to external data. Let’s compare them. First, Chainlink: Chainlink is not “one central server”. It’s a Decentralized Oracle Network (DON): •Multiple oracle nodes •Multiple data providers (exchanges, aggregators, etc.) •Offchain Reporting (OCR) → 1 aggregated value on-chain This is much better than a single API. But is it fully permissionless and trustless? Not really. Chainlink is curated: •Which oracle nodes are in a specific feed? Pre-selected. •Which data providers are trusted? Pre-selected. •What update thresholds / parameters are used? Pre-configured. You don’t just spin up a random node and join the main ETH/USD feed. So the Chainlink trust model looks like this: “We trust that a curated set of oracle operators •curated data providers •Chainlink governance and configuration will behave honestly and stay online.” Still decentralized vs a single server, but governance-centralized and permissioned at the edges. Now contrast that with the Kaspa oracle concept (as described by core researchers like Yonatan Sompolinsky): Kaspa already has: •A large set of PoW miners •A high-frequency blockDAG (10 blocks per second, aiming much higher) •A consensus model that samples the honest majority of hashpower in real time This is exactly what an oracle needs. In the Kaspa design, the oracle is not a separate network. There is no extra validator set. There is no separate DON. 👉 The oracle = the miners. Each block is not just “transactions”, it can also carry a vote on external data: price, event outcome, index inclusion, etc. How does that work in practice? •The protocol (via a standard) defines a question ID and type: •e.g. “BTC/USDT spot price for this 1s window” •or “Did Event X happen? YES/NO” •During a time window, each mined block includes a vote: •a price value, or •a boolean, etc. Each block = one PoW-weighted sample. Then every full node runs the same deterministic aggregation: •Collect all votes in the window •Discard obvious outliers •Compute e.g. the median •If dispersion is too high, mark as “unresolved” Result: a single oracle value that is: •Fully reproducible •Derived from the majority of hashpower •Computed by every node, not by a central aggregator So what’s the real decentralization difference? Chainlink: •Separate DON per feed •Node set: curated •Data sources: curated •Aggregation: done by the DON, then written as a single on-chain update Kaspa oracle concept: •No separate network •Oracle voters = the same PoW miners securing L1 •Aggregation: done by every node from raw on-chain votes In Chainlink, the final on-chain value is effectively: “This curated oracle committee signed off on this price.” In Kaspa, the final value is: “This is the deterministic aggregation of all votes emitted by the PoW majority in this time window.” Different trust anchors: •Chainlink → curated oracle set •Kaspa → raw hashpower majority Another key difference: real-time resolution. Kaspa’s blockDAG targets very high block rates (tens to hundreds of blocks per second). That means in ~1 internet round trip, you can sample dozens of independent miners, aggregate their votes, and reach a majority result with negligible probability of sampling a dishonest minority. That’s what Kaspa researchers call Real-Time Decentralization (RTD). In that model, PoW is not just “old-school security”. It becomes a high-frequency sampling mechanism: •Thousands of miners compete •Dozens of them win a block each second •Each win is a signed data point from that hashpower slice With enough samples, the honest majority emerges statistically, very fast.