TIDES · Institutional Intelligence Status current

Research first.
Build with proof.

The working record for TIDES’ institutional prospect-intelligence initiative: verified findings, reversed assumptions, build gates, and the shortest path to a validated first product.

Planning gateClosed
Gate 0 evidenced2 / 8
Research revisionsv0.4
Ops tasks complete14 / 17

Project timeline

August 28–29, 2026
America/Denver

AccessPublishing-path check created

An isolated marketing page was committed to confirm GitHub-to-Vercel access without modifying the production TIDES application.

v0.1Initial research and adversarial review

The first institutional prospect plan was challenged against SEC and Hyperliquid sources. Six early recommendations were reversed; mass ingestion was deferred in favor of a manual 50-account discovery slice.

v0.2Pricing hypothesis corrected

Public self-serve builder pricing clustered around 1–5 bps. A flat 10 bps fee was identified as a protocol maximum—not a normal institutional market rate—and likely hostile to high-volume adoption.

v0.3Unverifiable sample claim withdrawn

HyperTracker remained inaccessible, so the “710 builders / 1.4% sample” statement was withdrawn. A narrower 23-builder DefiLlama substitute was retained with explicit limitations.

v0.4Fee methodology resolved

Volume was correctly left unhalved and contaminated Insilico data was removed. Weighted observed fees revised from 4.34 to 4.73 bps; fee/volume correlation became −0.02, providing no meaningful relationship.

DecisionFurther desk research paused

No code was started. The team held the planning gate rather than convert unresolved pricing, onboarding, and customer assumptions into architecture.

RecoveryPublishing infrastructure repaired

Git attribution, pnpm, the marketing build contract, dedicated Vercel project, Git integration, Cloudflare DNS, TLS, automatic deployments, runtime checks, and visual evidence were completed. The timeline became publicly available at marketing.tides.market.

ExpansionOperating ledger introduced

The executive summary was expanded into a task-level operating record. Every task now carries a state, owner or dependency, acceptance evidence, and an explicit next transition.

ApprovalSlice 0 commercial hypothesis accepted

Jason approved subscription plus negotiable 1–2 bps usage pricing, the initial non-HFT target segments, and the discovery/risk/audit/reporting value thesis for the 50-account experiment. This is a test hypothesis, not permanent pricing policy.

Execution ledger

Infrastructure and operating tasks. A check means evidence exists—not merely that work was attempted.

✓ Done↻ In progress! Blocked· Queued

Publishing foundation

Repository → build → Vercel → DNS → public verification

10 / 10 complete
✓
Correct Git author identityOwner: automation

Repository commits now use the GitHub-recognized noreply address for the tidesmarket account.

Done
✓
Restore pinned package managerOwner: automation

Installed and verified pnpm 10.0.0, matching package.json.

Done
✓
Register marketing workspace appOwner: automation

Added package identity plus deterministic build, content test, and typecheck commands.

Done
✓
Configure Vercel outputOwner: automation

vercel.json builds with pnpm and publishes dist/.

Done
✓
Create dedicated Vercel projectOwner: automation

Project renamed to tides-marketing; Command Home remains a separate project.

Done
✓
Connect Git repositoryOwner: automation

tidesmarket/tides-product connected with apps/marketing-platform as root directory.

Done
✓
Attach custom domainOwner: automation

marketing.tides.market assigned to the production environment.

Done
✓
Create Cloudflare DNS recordOwner: automation

DNS-only CNAME marketing → Vercel’s assigned target; public resolver confirms it.

Done
✓
Verify TLS and public responseOwner: automation

HTTPS returns 200 with the expected TIDES timeline title and content.

Done
✓
Capture visual and runtime evidenceOwner: automation

Full-page browser capture completed; no page errors, and favicon 404 was eliminated.

Done

Platform reliability

Checks that support autonomous work over time

4 / 7 complete
✓
Marketing content testOwner: automation

Required timeline, Gate 0, v0.4, and 4.73 bps facts are asserted during test runs.

Done
✓
Marketing build testOwner: automation

Static artifact builds reproducibly into dist/index.html.

Done
✓
Automatic affected-app deploymentOwner: Vercel

Marketing changes deploy; unrelated Command Home builds correctly report “Skipped — Not affected.”

Done
✓
Repository synchronized and cleanOwner: automation

Working branch and main contain the publishing fixes with no uncommitted files.

Done
!
Repair memory indexingOwner: account administrator

Blocked: OpenAI embeddings return 429 billing_not_active. Reactivate billing or select another embeddings provider, then force rebuild.

Blocked
!
Restore full monorepo typecheckOwner: product engineering

Pre-existing strategy-lab-core error: TypeScript cannot resolve @tides/schemas. Marketing app checks remain green.

Blocked
·
Add automated freshness receiptOwner: automation

Next: expose deployment commit, last-updated timestamp, and verification time directly on the page.

Queued

Research audit trail

What changed, why it changed, and what evidence is still missing.

R-01 · REVERSED
“Charge the full 10 bps builder fee.”

The protocol maximum was incorrectly treated as the likely market price.

Current position

Validate subscription plus a negotiable 1–2 bps usage tier. Public self-serve observations cluster around 1–5 bps.

R-02 · REVERSED
“Largest 13F crypto-ETF holders are best first prospects.”

Size alone ignored fee tolerance and operating behavior.

Current position

Large holders often include market makers/HFT firms with near-zero fee tolerance. Start with crypto-native funds, RIAs, family offices, and treasury managers.

R-03 · PAUSED
“Build the large 13F database immediately.”

Architecture was getting ahead of offer and segment validation.

Current position

Run a manual 50-account Slice 0 first. Add ADV Part 2A intent extraction only after defining the target and legal addressability.

R-04 · WITHDRAWN
“The sample covers 1.4% of 710 builders.”

HyperTracker could not be independently accessed or its leaderboard verified.

Current position

Withdraw the claim. Retain the 23-builder DefiLlama substitute only as a bounded directional dataset.

R-05 · CORRECTED
“Observed weighted fee is 4.34 bps.”

Dataset contamination and volume-counting semantics required correction.

Current position

Do not halve the volume calculation; remove Insilico contamination. Revised weighted fee: 4.73 bps.

R-06 · UNPROVEN
“Lower fees reliably cause higher builder volume.”

The public sample is too narrow to establish this relationship.

Current position

Revised correlation is −0.02—effectively no demonstrated relationship. Treat pricing elasticity as a field-test question.

Gate 0 task plan

Business validation before architecture. Gate score is 2/8; remaining items require their own acceptance receipts.

Offer and addressability

Decisions and evidence required to open product planning

2 / 8 accepted
✓
G0-01 Freeze pricing hypothesisOwner: Jason · approved Aug 29, 8:21 AM MDT

Slice 0 offer: subscription/platform fee plus negotiable 1–2 bps usage pricing. Test evidence will determine exact subscription, minimums, and discount authority.

Accepted
✓
G0-02 Define differentiated valueOwner: Jason + product · approved Aug 29, 8:21 AM MDT

Working value: prospect discovery, portfolio/risk workflow, audit trail, reporting, and operational controls beyond Hyperliquid self-routing.

Accepted
!
G0-03 Approve segment matrixOwner: counsel

Accept when customer types × jurisdictions are labeled allowed, conditional, or prohibited with cited rationale.

Blocked
·
G0-04 Complete 50-account Slice 0Owner: research/sales

Accept with 50 verified entities, source provenance, buyer role, thesis, disqualifiers, and confidence score.

Queued
·
G0-05 Document data rightsOwner: counsel + data

Accept with allowed sources, terms, retention, refresh, deletion, and provenance rules.

Queued
·
G0-06 Model unit economicsOwner: finance/product

Accept with low/base/high routed volume, revenue, fee cost, data cost, support load, and gross margin by segment.

Queued
·
G0-07 Freeze vertical-slice acceptanceOwner: product + audit

Accept with inputs, outputs, held-out golden set, seeded defects, screenshots, and fail-closed rules.

Queued
!
G0-08 Record go/no-go decisionOwner: named executive

Accept with accountable owner, dated decision, evidence bundle, exceptions, and Gate 1 authorization.

Blocked

Slice 0 procedure

The manual 50-account experiment that determines whether software deserves to be built.

Discovery sample

Five batches of ten; review quality before expanding each batch

Rubric in progress · 0 / 50 verified
↻
S0-01 Define inclusion/exclusion rubricStarted Aug 29, 8:22 AM MDT · owner: automation

In progress: institution type, jurisdiction quarantine, AUM/holdings proxy, crypto activity, likely venue need, buyer role, confidence, and explicit disqualifiers.

In progress
·
S0-02 Select source mixEvidence design

13F for holdings signals; ADV Part 2A for strategy intent; websites, filings, and primary disclosures for verification.

Queued
·
S0-03 Research accounts 1–10Quality calibration

Dual-check entity identity, source date, buyer role, reason-to-believe, and legal status before continuing.

Queued
·
S0-04 Research accounts 11–50Four gated batches

Proceed only if prior batch meets completeness and false-positive thresholds.

Queued
·
S0-05 Score without fabricated precisionAnalysis

Separate observed facts from inference; retain source links, confidence, missing fields, and disqualifiers.

Queued
·
S0-06 Test messaging and objectionsSales validation

Track replies, qualified conversations, price objections, value objections, onboarding blockers, and false positives.

Queued
·
S0-07 Publish decision memoGate receipt

Recommend build, revise, or stop with sample evidence and explicit confidence limits.

Queued

Gate 1 build outline

Queued—not authorized. This becomes active only after Gate 0 passes.

First vertical slice

One inspectable path from source evidence to prioritized prospect

0 / 9 started
·
G1-01 Source registry

Canonical source IDs, licenses/terms, fetch receipts, timestamps, hashes, and provenance.

Gated
·
G1-02 Entity resolution

Deterministic identifiers, aliases, parent/subsidiary relationships, collision review, and manual overrides.

Gated
·
G1-03 Filing and intent extraction

Observed passages preserved beside structured holdings, strategy themes, and confidence.

Gated
·
G1-04 Regulatory hard gate

Prospects cannot score or enter outreach queues until addressability passes.

Gated
·
G1-05 Explainable scoring

Feature values, weights, missingness, exclusions, and score explanation visible per account.

Gated
·
G1-06 Review queue

Human approval, rejection reasons, duplicate merge, source drill-down, and audit history.

Gated
·
G1-07 Held-out golden set

Known valid/invalid entities and extraction results excluded from prompt and implementation tuning.

Gated
·
G1-08 Seeded defect gauntlet

Wrong entity, stale filing, prohibited segment, missing provenance, duplicate, and fabricated evidence must fail closed.

Gated
·
G1-09 Deployment evidence bundle

Green tests, screenshots, console/network checks, commit/deployment receipts, rollback path, and auditor verdict.

Gated

Gate 0

What must be true before product architecture is frozen.

Addressable segmentsCounsel must mark customer types and jurisdictions allowed, conditional, or prohibited.
50-account discoveryManually source, score, and test outreach assumptions before mass ingestion.
Data rightsDocument permitted sources, retention, provenance, and refresh expectations.
Volume economicsModel plausible routed volume and cost-to-serve by target segment.
Acceptance criteriaFreeze a testable vertical-slice definition with held-out evidence.
Go / no-go ownerName the accountable decision maker and record the gate outcome.

Recommended next move

A bounded path that buys evidence before infrastructure.