LITE-PSM-USDC-A Parameter Change

TLDR BA Labs, acting on behalf of the Core Council, proposes the following parameter changes to the LITE-PSM-USDC-A: Increase LITE-PSM-USDC-A buf from 400M to 800M Increase LITE-PSM-USDC-A gap from 400M to 800M Context For a complete overview of the technical scope of the LITE-PSM-USDC-A, refer to this forum post: LitePSM (LITE-PSM-USDC-A) : Introduction and Overview The migration from PSM-USDC-A to the LITE-PSM-USDC-A has been thoroughly monitored through different analyses, refer to the following forum posts: LitePSM Migration Period Analyses The previous parameter changes can be found in the following forum posts: LITE-PSM-USDC-A Phase 1 (Test Period): Proposed Parameters LITE-PSM-USDC-A Phase 2 (Major Migration): Proposed Parameters LITE-PSM-USDC-A Phase 3 (Final Migration): Proposed Parameters LITE-PSM-USDC-A has been live for ~23 months and has settled into the role of the dominant USDC ↔ DAI venue inside the Sky stablecoin system. The contract has processed 1,027,167 swap events since deployment, with current USDC reserves at 4.13 B against a buf of 400 M. The last buf recalibration was performed on 2024-10-07 (phase 3), when reserves stood at ~1.98 B. Since then USDC reserves have more than doubled (+108%) while liquidity-buffer parameters have remained flat. Reserves first reached the ~4 B level in April 2025 and have hovered at or above it since, yet buf and gap have stayed at 400 M throughout making a recalibration overdue. This post proposes a methodology to size buf and gap using on-chain history of swap sizes, daily flows, and reserve trajectory. On buf and gap The LitePSM holds two independent buckets, and the two swap directions draw on different ones: Refresh capacity A SellGem swap is served instantly from the pre-minted DAI bucket (buf). As that bucket depletes, the fill keeper mints fresh DAI to top it back up but minting is capped by the ilk’s debt ceiling (line), and the DC-IAM raises line by at most gap every ttl. With ttl = 12 h, the line can step up twice per day, so: Refresh capacity = 2 × gap per day: the rate at which the buffer can be replenished, i.e. the sustained daily SellGem throughput the PSM can support before it runs dry. Single-day capacity = buf + 2 × gap: a single day, starting from a full buffer: the whole bucket plus that day’s two refills. At the current buf = gap = 400 M, single-day capacity is 0.4 B + 0.8 B = 1.2 B; the heaviest observed SellGem day (1.75 B) exceeds it. At the proposed buf = gap = 800 M, single-day capacity rises to 0.8 B + 1.6 B = 2.4 B, clearing every day on record. Empirical observations USDC reserves trajectory The buf/reserve ratio at the last intentional sizing decision (phase 3) was ~20 %. USDC reserve growth since then has eroded that ratio. Single-swap distribution (SellGem) Because only the SellGem direction is constrained by buf, the relevant tail is the largest SellGem swaps. Only 2 SellGem swaps in ~23 months have exceeded the current 400 M buf. Daily flow distribution Last-90-day distribution of daily one-way volumes: The most stressful single SellGem day in the last 90 days drained ~1.75 B of DAI. With current gap = 400 M and ttl = 12 h, the line-refresh capacity is only 2 × gap = 800 M / day which is below even the P90 of recent daily SellGem volume, and far below the heaviest days. Largest single-day reserve drawdowns The metric that gap must keep pace with is the largest single-day SellGem volume. The gross amount of pre-minted DAI withdrawn in a day, which fill must replenish: 1.75 B (2026-05-18) 1.60 B (2025-06-20) 1.41 B (2025-10-21) 1.41 B (2026-03-05) 1.31 B (2026-01-13) All five exceed the current 800 M / day refresh capacity (2 × gap at gap = 400 M). At the proposed gap = 800 M, refresh capacity rises to 1.6 B / day, and combined same-day serving capacity (buf + 2 × gap = 2.4 B) covers every day on record. Methodology We anchor the new buf & gap to three independent sizing lenses and recommend the maximum, since each captures a different failure mode. The different methodologies allow to keep continuity with prior governance reasoning (A), ensure the main UX promise of the LitePSM (gas-cheap pre-minted swaps) holds even on tail trades (B), and ensures the DC-IAM mechanism does not become the binding constraint on legitimate flow (C). Method A - Reserve proportional Replicate the 20.2 % buf : reserve ratio that BA Labs implicitly endorsed at the phase-3 recalibration (the last time buf was set as a deliberate function of reserve, rather than left flat). buf_A = 0.202 × target_reserve Method B - Largest historical swap coverage Size buf so that the largest historically observed single swap, scaled into the new reserve regime, can be served from pre-minted USDS without on-the-fly minting. The largest SellGem on record (543 M) was 14.4 % of contemporaneous reserves. buf_B = 0.144 × target_reserve Method C - Daily flow throughput buf and gap together must let the line refresh fast enough to absorb a heavy one-way day without throttling. With ttl = 12 h, daily refresh capacity = 2 × gap. We require this capacity to cover the recent distribution of max one-way daily volume (max(buy, sell) per day), scaled to the new reserve regime. gap_C ≥ (P_x_max_oneway × scale_factor) / 2 where scale_factor = target_reserve / current_reserve. Using max-one-way rather than buy-only makes the test direction-agnostic: a heavy sell day stresses the system as much as a heavy buy day. Results Recommendation for 4 B reserve target Proposed: buf = gap = 800 M, line = 10 B, ttl = 12 h Recommendation for 5 B reserve target Proposed: buf = gap = 1B, line = 10 B, ttl = 12 h Parameter Recommendation BA Labs, in our capacity as Core Council Risk Advisor, recommends the following parameter changes to the Core Facilitators and include them in the next Atlas edit, modifying A.3.3.2.7.1.1.2 - Lite Peg Stability Module Parameter Values: buf: Set to 800,000,000 DC-IAM gap: Set to 800,000,000 Atlas Authorization The recommendation must be approved by the Core Facilitator. CC: @Jansky @ldr 1 post - 1 participant Read full topic