Roadmap

Phase I — Runtime expansion

Expanding Protean's bounded peptide-search runtime from simulation-scale orchestration into stronger local structure support, more durable scientific memory, and review-gated assay preparation.

The first $30,000 in WETH-side bankr fees is allocated against three concrete infrastructure-activation targets. The framing is protocol activity expands scientific capability — capability earned by the system's own throughput, not promised by a roadmap line.

Allocation 1 — 1× NVIDIA RTX PRO 6000 Blackwell · 96 GB

Purpose. Local CUDA structure-generation and folding workloads, including RFdiffusion, BindCraft, AlphaFold-class validation, Boltz/Chai-style model workflows, and future GPU-only protein design tools. Continuous on-prem execution ends the cycle's dependency on staged or cloud-bound structure passes.

Allocation 2 — WRX90 platform + 256 GB ECC RAM

Purpose. Stable continuous runtime operation for Galen cognition loops, ingestion, scoring, indexing, local model execution, provenance preparation, and long-running scientific orchestration. Error-corrected memory matters for runtimes that hold cycle state across days.

Allocation 3 — First review-gated assay preparation batch

Purpose. Move top-ranked computational candidates into human-reviewed assay preparation so Protean can begin connecting bounded candidate generation, biological evidence, and experimental feedback. Review-gated; no validation claim exists until real assay data is returned and reviewed.

Runtime impact

• Continuous local RFdiffusion execution
• Concurrent folding + validation pipelines
• Persistent Galen cognition cycles
• Higher candidate throughput
• First review-gated assay preparation batches
• Replayable experimental lineage
• Stronger failure learning from assay outcomes

Shipped

The trust architecture is live. None of the items below are aspirational.

• Protean Ledger on Base mainnet. UUPS-upgradeable contract at 0xE3c261F3C05D4c4710003cd6066EfD95094cf5f0, schema protean.ledger.v1, 17 RecordType enum values, 20 RelationType enum values, 10 LifecycleState values, 6 DisclosureState values.
• Three-principal role separation. Treasury (DEFAULT_ADMIN_ROLE, UPGRADER_ROLE, PAUSER_ROLE, RETRACTOR_ROLE, LINEAGE_REVOKER_ROLE, IP_DECLARANT_ROLE), operator (OPERATOR_WRITER_ROLE and operational roles where granted), Bankr (AUTOMATION_WRITER_ROLE only). Operator compromise is not governance compromise.
• Operator approval-token gate. Single-use 5-minute-TTL tokens at ~/.openclaw/exec-approvals.json (mode 0600). Bankr refuses to broadcast without an unexpired token bound to the actor + capability + review-record-id.
• Public explorer at protean.sh/ledger. Per-record pages with chain-anchored provenance strips, lineage walks, and "Verify this record yourself" copy-paste commands.
• Open-source indexer. scripts/index_ledger_from_genesis.py, reproducible state digest at /ledger/api/v1/indexer/digest. Any third party can recompute it from the event log alone.
• Public verification mirrors. github.com/proteanlabs1/ledger-mirror and the Gitlawb repos under DID did:key:z6Mkt6MEeSCJM2krT1PfX8BmTWbi9YYkLqdaRXSF6UZvy5QB carry public verification artifacts. The mirrors are supplemental; the chain and Digest remain authoritative.
• ScientificAsset and IPAsset records. First-class on-chain record types for citable aggregations and provisional-IP intent. IPAsset minting is treasury-gated.
• Scheduled dry-lab orchestration with supplemental snapshots. Constraint-guided candidate proposal, failure-aware ranking with seven canonical weights, bounded feedback when trusted preconditions exist. Snapshots are local integrity scaffolding downstream of the on-chain record.
• Live hourly treasury snapshot. No live RPC at page render; cron-captured into Neon every hour.

Next

The closest closable gaps are operational, not architectural.

• Artifact automation. Auto-author the per-record mirror_sources/artifacts/mainnet/<recordId>.json from chain state when a new record is confirmed. Removes the bookkeeping toil that today keeps post-bootstrap records carrying placeholder replay anchors.
• Mirror auto-publish on chain event. A repository_dispatch from the indexer cron fires the publish workflow within minutes of a new confirmed record. Removes the "next Bio-repo push" lag.
• Indexer-health Galen cron. Hourly observation that the live indexer digest stays in step with the chain head. Telegram alert on degraded.
• CRON_SECRET enforcement. Tighten the Vercel cron auth so unauthenticated POSTs cannot force a re-index.
• Per-step failure_policy and review_gate enforcement inside loop_runner.py.
• Contradiction graph as a default cycle stage with explicit memory consolidation.

Then

• Programmable assay-preparation handoff: simulated dry run, provider adapter review, payment integration on Base, provider packets as typed Ledger edges.
• Remaining 15 wet-lab lifecycle states (scientific_review_pending through closed).
• Collection-evolution + wet-lab outcome memory back into the optimisation loop.
• Treasury multisig migration once on-chain spend cadence justifies it.
• Reviewed-public-collection release workflow as Ledger Collection records.

Reserved

Network-aligned coordination research over public-safe state — reviewed collections, attestation records, contribution receipts. Reserved design space, not a committed mechanism. Any future structure would be separately constituted, separately documented, and subject to legal, regulatory, jurisdictional, compliance, scientific, tax, and operational review.

The roadmap is direction, not date. The criterion for moving an item from one phase to the next is that the relevant control surface has been strengthened — never weakened — by the change.

Boundary

Every roadmap item respects the safety contract in Scientific boundaries. Expansion that would require relaxing a refusal site is, by construction, not on the roadmap.