PodBrowser
The Edge Podcast

Don't Touch A Stablecoin Without Checking Pharos First | DeFi Frontier

Thursday, 20 August 2026 · 4 min read · Listen to the episode ↗

In this episode, the founders of Pharos explain how their stablecoin risk monitoring platform aggregates reserve composition, exit liquidity, minting authority, and peg resilience into a single safety score updated every two hours across roughly one hundred sixty stablecoins.

Pharos is a stablecoin risk monitoring platform built by Bryce, originally as a personal offline dashboard to replace the fragmented workflow of managing Dune, DeFi Llama, CoinGecko, and protocol documentation across multiple tabs. AI made the build possible because Bryce was not originally a developer. Ike joined after winning a three-thousand-dollar contest Bryce ran to drive awareness, with his entry receiving roughly twenty thousand views. The day after their first conversation, Ike presented a full PowerPoint outlining a more ambitious vision than Bryce had originally planned, and the contest prize was treated as a signing bonus.

The core problem Pharos addresses is that most stablecoin users evaluate only peg stability and APY rather than the safety of underlying collateral layers. With approximately one hundred sixty stablecoins in existence, manual tracking of reserve composition and risk profiles is impractical. Pharos updates reserve data every two hours by reading on-chain records and has logged approximately sixty-seven thousand depeg events across all tracked stablecoins. Traditional analytics platforms typically produce a single point-in-time report rather than continuous monitoring, which Pharos treats as a meaningful gap.

Pharos reserve data reveals some counterintuitive compositions. USDT is dominated by T-bills with smaller allocations of physical gold, bitcoin, and non-T-bill cash, with those non-T-bill components representing roughly ten to twelve percent each. USDC reserves consist entirely of bank deposits in different forms. USDe is now approximately ninety-nine percent composed of other stablecoins and only one percent basis trade, meaning the basis trade backing it was originally described as relying on has not been the case for roughly a year. The platform uses color coding so that orange and red in a reserve visualization signals lower-grade collateral.

The Pharos safety score aggregates reserve composition, exit liquidity, peg resilience, minting authority, dependency risk, collateral quality, bridging model risk, and peg stability into a single metric. Minting authority is treated as a frequent attack vector. USDD scored ten out of one hundred on minting authority. MIM by Abracadabra has had its minting setup abused three times and is currently trading at approximately twelve cents with no recovery path identified. CRVUSD scores sixty-three out of one hundred on minting authority because governance allows veCRV votes to mint unbacked crvUSD for certain purposes, constituting a potential attack vector if governance were compromised. BOLD from Liquity receives a clean minting authority rating because the only authorized minter is an immutable smart contract requiring appropriate collateral. Liquity Bold is the highest-graded stablecoin on the platform under the current scoring system. Stablecoins currently rated A include WBTC, LUSD, and sDAI. The team stated they do not hold significant capital below an A or B rating.

The PM USD case became a defining credibility moment for Pharos. Pharos rated PM USD a D before its depeg. The stablecoin had no redemption mechanism despite reaching approximately one hundred million dollars in market cap, marketed itself as gold-backed by in-situ unmined gold held in no vault, and had real decentralized exchange liquidity of only approximately three hundred seventy-five thousand dollars against that supply. The project used Curve bribes to make its pool appear healthier than it was. Community questions about collateral were deflected on Discord, and a direct call confirmed that a claimed partnership with Aptos Max and Kitco was never in place. The PM USD team messaged Pharos objecting to the low rating before the depeg, and after publication one individual connected to the project attempted to involve a European regulator against the Pharos analyst. After the depeg, a user messaged the team saying the Pharos report helped them exit in time.

The current scoring system can produce a B rating for a stablecoin with unsafe collateral if strong liquidity and low dependency scores compensate for a poor resilience score. Safety score version nine will introduce threshold logic that forces the overall score down when any critical dimension crosses a defined threshold, preventing that compensation effect. Version nine will also be mechanism-aware, applying branching logic based on stablecoin type. For centralized real-world-asset stablecoins, MiCA authorization will factor into the score. For maximally decentralized immutable stablecoins, MiCA compliance is excluded because regulatory bodies have no practical enforcement power over them. Decentralization is being removed as a weighted scoring factor in the next version but retained as an informational signal, after which USDC could potentially achieve an A rating.

Pharos is approximately four months old with a monthly burn rate of roughly seventeen hundred to eighteen hundred dollars. A Gitcoin round covered about two months of operating expenses. The general dashboard will remain free, with monetization through paid API keys and services sold to DAOs and protocols. Pharos is in discussions with Pendle and Compound to integrate its safety scoring directly into their markets. The Telegram bot, described as the most popular feature, allows custom alerts for depegs exceeding one hundred basis points, safety score changes, and reserve composition changes. The platform also maintains a cemetery ledger tracking all dead stablecoins with causes of death recorded by category, and case studies of both stablecoins that failed and those that survived.

This summary was generated from the episode transcript and can contain mistakes.