fix: repair broken PoC links in writeups

src/posts/2026-02-05-private-bonds-on-privacy-l2s.md

@@ -216,6 +216,6 @@ The tradeoffs are also structural to this approach:
 
 For teams that want to ship a prototype without building cryptographic infrastructure from scratch, privacy L2s offer a faster starting point. For teams that need precise control over every layer, the custom UTXO approach from Part 1 remains viable.
 
-The full implementation (on Aztec) is [open source](https://github.com/ethsystems/pocs/tree/main/pocs/private-bond/privacy-l2), with a detailed [specification](https://github.com/ethsystems/pocs/blob/main/pocs/private-bond/privacy-l2/SPEC.md) covering the protocol design.
+The full implementation (on Aztec) is [open source](https://github.com/ethsystems/pocs/tree/master/pocs/private-bond/privacy-l2), with a detailed [specification](https://github.com/ethsystems/pocs/blob/master/pocs/private-bond/privacy-l2/SPEC.md) covering the protocol design.
 
 In Part 3, we will explore a third approach: fully homomorphic encryption (FHE). Where UTXO models hide data by never putting it on-chain, FHE allows computation on encrypted data directly. Different cryptography, different tradeoffs, same institutional requirements.

src/posts/2026-02-19-building-private-transfers-on-ethereum.md

@@ -22,7 +22,7 @@ Traditional banking solved this decades ago. Payment details are visible only to
 
 In a [previous post](/building-private-bonds-on-ethereum/), we built private zero-coupon bonds using a UTXO model and ZK proofs. That PoC demonstrated the cryptographic primitives: commitments, nullifiers, Merkle trees, encrypted memos. This one tackles a different problem: stablecoin payments where compliance gating, not just privacy, is the primary design constraint.
 
-This post walks through a proof-of-concept that brings banking-grade payment privacy to stablecoin transfers on Ethereum L1. The design prioritizes compliance-first privacy: only KYC-verified participants can enter the system, and viewing keys enable selective disclosure for regulators. The full implementation is [open source](https://github.com/ethsystems/pocs/pull/15), with a detailed [specification](https://github.com/ethsystems/pocs/tree/main/pocs/private-payment/shielded-pool/SPEC.md).
+This post walks through a proof-of-concept that brings banking-grade payment privacy to stablecoin transfers on Ethereum L1. The design prioritizes compliance-first privacy: only KYC-verified participants can enter the system, and viewing keys enable selective disclosure for regulators. The full implementation is [open source](https://github.com/ethsystems/pocs/pull/15), with a detailed [specification](https://github.com/ethsystems/pocs/blob/master/pocs/private-payment/shielded-pool/SPEC.md).
 
 ## The Gated Shielded Pool
 

src/posts/2026-03-05-private-crosschain-atomic-swap-part-1.md

@@ -146,7 +146,7 @@ ZK circuits verify note formation and ownership. Shielded pools prevent double-s
 
 The coordinator is the only component not yet specified. It could be built from a Trusted Execution Environment, a multi-party computation protocol, or fully homomorphic encryption, each with different trust assumptions and performance trade-offs. In Part 2, we pick one: a TEE running in AWS Nitro Enclaves. We go inside the enclave, examine what attestation actually proves, work through the real attack surfaces, and walk through what the demo logs show.
 
-The full implementation is open source, with a detailed [specification](https://github.com/ethsystems/pocs/tree/main/pocs/approach-private-trade-settlement/tee_swap/SPEC.md) and an [interactive protocol walkthrough](/tee-protocol-page).
+The full implementation is open source, with a detailed [specification](https://github.com/ethsystems/pocs/blob/master/pocs/private-trade-settlement/tee_swap/SPEC.md) and an [interactive protocol walkthrough](/tee-protocol-page).
 
 ## References
 

src/posts/2026-03-18-private-crosschain-atomic-swap-part-2.md

@@ -147,7 +147,7 @@ The TEE coordinator is a starting point, not the destination. MPC could replace
 
 The coordination problem reduces to this: two parties each hold private inputs (their ephemeral key and encrypted salt), and we need a single proof that both sets of inputs are consistent with the on-chain state. That is what co-SNARKs solve — each party contributes their secret inputs to a joint ZK proof without revealing them to anyone. The proof itself becomes the atomic revelation. If it verifies, both sides are consistent. No trusted intermediary, no hardware assumption, no coordinator to compromise. The coordinator becomes a protocol rather than a party.
 
-The full implementation is open source, with a detailed [specification](https://github.com/ethsystems/pocs/tree/main/pocs/approach-private-trade-settlement/tee_swap/SPEC.md) and an [interactive protocol walkthrough](/tee-protocol-page).
+The full implementation is open source, with a detailed [specification](https://github.com/ethsystems/pocs/blob/master/pocs/private-trade-settlement/tee_swap/SPEC.md) and an [interactive protocol walkthrough](/tee-protocol-page).
 
 ## References