Independent risk research · Singapore
The independent risk layer for onchain dollars.
Yield-bearing dollars, tokenized funds, and DeFi vaults are landing on institutional books faster than anyone can vet them. Arunights grades their loss resilience on one open scale — every input evidence-linked, confidence and falsifiers published, and no issuer ever pays for a grade.
The gap
Onchain dollars are landing on institutional books faster than the controls around them.
A committee can see the yield. What it cannot see is whether the backing behind a token holds up when redemptions spike, an oracle goes stale, or a counterparty stops paying. That is where allocations stall.
Trustees, risk committees, and supervisors all want to know who checked the instrument, what evidence the read stands on, and who is watching it once it is live. Those are the questions Arunights answers — in public, with every input linked.
What Arunights does
One universal scale across three outputs that share one record: a published grade, collateral monitoring on a live book, and a reproducible audit evidence pack. The grades are open to everyone. Monitoring runs with design partners.
Risk Grades
23 graded · draft/provisionalAn independent grade of an onchain dollar instrument, vault, or curator on one universal loss-resilience scale, A+ to F. Six class-specific pillars, weighted, with structural flaws that cap the grade regardless of score — written up with the evidence behind every pillar, a confidence label, and the falsifier that would change it.
View the gradesCollateral Monitoring
Design-partner previewOnce an instrument is live, grade-derived haircuts and policy checks — concentration, fragile-tier and low-confidence exposure — run against a book, and a flag is raised the moment a threshold is crossed. Every flag carries its metric and its source. An analyst reviews each one; agents never move a grade.
Open the monitorAudit Evidence
ReproducibleEvery review is a hash-verified evidence pack: two sha256-pinned inputs and a reproducible manifest hash, so any counterparty, auditor, or regulator can re-derive every figure without trusting us.
See a sample packHow we grade risk
Collateral fails at its weakest point, not at its average. So we score six class-specific pillars on their own, weight them, and then let specific structural flaws cap the grade regardless of the weighted score — an unverifiable-reserves or undisclosed-external-manager finding overrides a comfortable average. Nothing settles into a comfortable middle.
Backing & reserves
What actually backs the token, how often it is attested and by whom, and whether the backing is exogenous (T-bills) or endogenous (its own ecosystem assets).
Redemption & exit
Direct redemption rights and their gates, cooldowns, and minimums, plus secondary-market depth — and how both behaved the last time exits spiked.
Counterparty & custody
The custodians, exchanges, and external managers in the loop. A discretionary off-chain manager is the deadliest single flag — the Stream lesson.
Peg & market infrastructure
How lending markets price it — a hardcoded $1.00 oracle amplifies contagion — plus depeg history and liquidity fragmentation.
Leverage & rehypothecation
How much of supply is looped as collateral elsewhere, and whether the asset is a load-bearing block in a leverage tower — the xUSD pattern.
Transparency & governance
A named legal entity and team, documentation quality, admin keys, upgradability, and timelocks.
Six pillars shown for yield-bearing dollars; curated vaults, curators, and tokenized-RWA wrappers each carry their own six, weighted for the class.
Software does the watching. A named human signs every score, haircut, and grade before it ships. Monitoring agents surface flags for review — they never change a grade on their own.
The method, applied in public
Before you trust the method on your own book, watch it work on someone else’s. We took two real failures apart in public and tagged every material claim to its source.
The credit you could not see
Why on-chain repayment flows never revealed a $36M off-chain credit failure at Maple Finance. A provenance-tagged teardown.
The Stream Finance collapse
How a yield-bearing synthetic dollar lost its peg, and why four of its five risk channels were flagged in public first.
Who works with us
Issuers and curators
You are bringing an onchain dollar or pool to market, and your first institutional buyer will want an outside read on it. Better to have one ready.
Desks and allocators
Capital is going into onchain dollars and vaults, and the committee wants to know where the structure is thin while the allocation is still a decision.
Risk and compliance
You answer to a board or a supervisor, and you need a record of what was checked, by whom, and what has moved since.
Who runs it
Dhruv works on counterparty and collateral risk in tokenized real-world-asset markets. The methodology and the monitoring framework both come out of that research — every number the engine produces traces back to something written down and evidence-linked, not asserted.
He also founded JAYIC, a national pre-university investing competition run with JA Singapore and HSBC.