diff --git a/DLT_EVALUATION.md b/DLT_EVALUATION.md
index 956ebe7..54f6241 100644
--- a/DLT_EVALUATION.md
+++ b/DLT_EVALUATION.md
@@ -9,7 +9,9 @@ Per [SPEC §13.1](./SPEC.md#13-dlt-requirements--profiles), plus project policy,
 1. **Sub-second finality** — real, deterministic finality (BFT consensus commit) under ~1 s on mainnet *today*. Optimistic preconfirmations and centralized-sequencer "soft confirmations" do not count.
 2. **Very low per-notarization cost** — a small contract-call write (an ANP anchor: hash + ~200 bytes metadata) well under $0.01, predictably.
 
-Additional graded criteria: programmable escrow with time-window logic and **contract-readable anchor state** (the §6.2.1 uncontested path must be able to prove the *absence* of a dispute on-chain — see issue [#6](https://github.com/byte5ai/anp/issues/6)), native on-chain randomness (VRF arbiter/witness draws, [#9](https://github.com/byte5ai/anp/issues/9)), sponsored transactions, native stablecoins, maturity. **Quantum resistance is reported separately for every chain** — it is a graded criterion, not a hard filter, because no candidate chain offers post-quantum account cryptography today; ANP's binding layer is PQC off-chain by design (§6.5).
+Beyond those two discriminating filters, SPEC §13.1 also **mandates** (MUST) **programmable smart contracts** (req 3), **programmatic wallets** (req 4), **contract-readable dispute/enforcement state** (req 6 — the §6.2.1 uncontested path must be able to prove the *absence* of a dispute on-chain, see issue [#6](https://github.com/byte5ai/anp/issues/6)), and **anchor discoverability** (req 7). The first two are table-stakes for any smart-contract chain; **contract-readable anchor state** is the MUST that actually disqualifies otherwise-cheap, fast anchoring layers — it is why Hedera HCS-only anchoring fails (see the Hedera row below). These are pass/fail, not graded; they simply don't *discriminate* among the shortlist because every shortlisted chain clears them.
+
+Genuinely **graded** criteria (scored, not pass/fail): native on-chain randomness (REQUIRED only where the profile uses VRF arbiter/witness draws, [#9](https://github.com/byte5ai/anp/issues/9); graded otherwise), sponsored transactions, native stablecoins, maturity. **Quantum resistance is reported separately for every chain** — it is a graded criterion, not a hard filter, because no candidate chain offers post-quantum account cryptography today; ANP's binding layer is PQC off-chain by design (§6.5).
 
 Findings are source-backed (research notes with URLs are preserved in the repository issues and PR descriptions); numbers flagged *vendor-sourced* or *inferred* were not independently measurable.