Colosseum · Solana Frontier · 2026 v0.2 · May 2026
Soothsayer.
Oracle-aware liquidity that shows its math.
An oracle-aware spot AMM for tokenized RWAs on Solana, built on Soothsayer's calibrated fair-value band. Soothsayer Research + Oracle-aware AMM.
Colosseum hackathon · 12 slides Adam Noonan — 2× exited founder

Hook · the gap

Every weekend, $3.5B of tokenized equities on Solana have no principled price.

MonTueWedThuFriSatSun
ChainlinkLTP frozen / tokenized mark live
Pyth + Blue Ocean24/5 overnight ATS
Soothsayerfair-value + CI
The 48-hour gap — Fri 8pm → Sun 8pm ET
— Chainlink weekend: frozen LTP, placeholder-derived v11 mid
— Consequence: CPMM LPs eat the Monday-open repricing as LVR · lenders carry the same gap as bad-debt risk
Soothsayer.
02 / 12

Why now.

Market · April 2026
$25B+
xStocks cumulative volume
95–99% of all tokenized-equity trading activity lives on Solana.
2×/mo
Kamino xStock supply
$25M total supply (Q1 2026 snapshot); supply caps doubled across 6 tickers in March 2026.
$26.4B
Tokenized RWA onchain · +300%
RWA is the fastest-growing segment of onchain finance entering 2026.
Incumbent admission
“Without these capabilities, tokenized products are exposed to potential mispricings, unfair liquidations, regulatory gaps, and other critical risks.”
Chainlink · v10/v11 launch blog · blog.chain.link
Soothsayer.
03 / 12

Why this matters — the used-car-lot analogy.

ELI5
01 · Normal day
Big auction open.
Auction price board $42,800
12 active buyers · Deep book
Fair price = auction + small dealer markup. Market is deep, competitive, honest.
02 · Saturday, rainstorm
Auction closed.
Auction price board — CLOSED —
1 trade · $21,400 · –10 on the print
One trade in an empty lot. Technically “the market price” — but it’s not what the car is worth.
03 · Soothsayer reading
Fair value + honest uncertainty.
FV · CI
95% CI
$41,900
Fri 20:00SatSun 20:00
Friday close ES=F return VIX regime label
Soothsayer publishes fair value + honest uncertainty. Protocols don’t liquidate on a print from an empty lot.
Open market ≠ fair price when nobody’s around. Soothsayer is the model that tells you which is which.
Soothsayer.
04 / 12

The coverage gap.

Competitive matrix
Provider Weekend coverage Methodology Confidence interval Open source
Chainlink Data Streams ~ field-dependent Frozen LTP or opaque tokenized mark — none verifiable — closed
Pyth + Blue Ocean — overnight only Weighted median 10.2% at claimed 95% ~ partial
RedStone Live ~ 24/7 marketed ** — undisclosed — none on public feed ** — closed
Soothsayer ✓ weekend + CI ✓ published math ✓ aggregate-feed receipts ✓ Apache-2.0
* Pyth's documentation recommends per-publisher self-attestation to ~95% coverage, but the aggregate served feed downstream protocols actually read does not publish a verifiable calibration receipt. Soothsayer publishes receipts at the aggregate-feed level — replayable against the persisted surface and 12 years of public data.
** Measured against RedStone's public REST gateway (api.redstone.finance/prices, 2026-04-26): point-only schema with no confidence field, 30-day retention cap, partial equity coverage (SPY, QQQ, MSTR available; TSLA, NVDA, HOOD, GOOGL empty; underliers only — no SPL xStock tokens). RedStone's paid Live tier may serve a different stream; not verifiable from public artifacts.
Wins 4 of 4 axes Coverage is the race everyone’s running. Calibration transparency is the one nobody’s entered.
Soothsayer.
05 / 12

The product stack

Soothsayer Research, and the oracle-aware AMM it makes possible.

01 · The moat
Soothsayer Research — the only auditable fair-value band on Solana.
Papers 1 and 3 ship the calibrated band: a fair-value interval whose realized coverage is empirically verifiable on 12 years of public data. OOS Kupiec + Christoffersen pass at τ ∈ {0.68, 0.85, 0.95}. Anyone can fork an AMM. Only Soothsayer ships an auditable fair-value receipt under it.
02 · The lead product
Oracle-aware spot AMM — liquidity that knows its uncertainty.
A spot AMM for tokenized RWAs that consumes the band on-chain. Execution decisions reference fair value and its calibrated bounds, not just the on-chain mid — so LPs survive Monday opens, halts, and oracle drift. Devnet MVP this hackathon; mainnet on a single xStock pair after.
03 · The future stack
Perp, options, and settlement — same receipt.
Every downstream primitive needs uncertainty as an input. With the band on-chain and the AMM live, perp mark/liquidation logic, calibrated options vol, and licensed band feeds to third-party venues all reuse the same calibration contract.
Stack frame: docs/product-stack.md · L0 oracle (live) → L1 AMM (this hackathon) → L2 perp / L3 options / L4 settlement.
Soothsayer.
06 / 12

The lead product · architecture

An AMM that prices like the underlying still trades — because the band tells it where fair value is.

01 · FEED
Calibrated band, on-chain.
Soothsayer Research publishes point, lower, upper, plus regime + forecaster receipt. Rust at byte-parity with the Python reference (75/75); router on devnet since 2026-04-29.
Layer 0 of the product stack.
02 · MECHANISM
Band-aware Anchor program.
Each block, the AMM reads the band and conditions execution on it — so quoting, spread, fees, or LP gating respond to fair-value drift before arbs do. The band is the LP's stop-loss; the receipt is the LP's audit trail.
Devnet MVP — single xStock pair.
03 · OUTCOME
LPs survive the windows that bleed CPMMs.
Weekend repricing into Monday open, halt-window mid-drift, primary-oracle drift, vol regime shifts. Every CPMM LP eats these as LVR. The oracle-aware AMM repositions before the toxic flow arrives — and publishes a receipt for why.
Future stack: same band, same receipt — perp / options / settlement.
MVP mechanism [TBD pending implementation brief]  — one of: band-conditioned fee · band-conditioned spread · JIT-LP gating · band-breach auction. Pick one for the devnet demo; the architecture above is uniform across choices.
Soothsayer.
07 / 12

Five windows that bleed CPMMs. The oracle-aware AMM survives them.

LP scenarios
01
Weekend LVR
CPMM
Pool mid stays at Friday close. Underlying drifts on ES futures + crypto for 48 hours. LPs eat the difference Monday morning.
Oracle-aware
AMM tracks the band over the weekend and repositions quote / spread before the open. Toxic flow finds a different pool.
02
Monday-open arb sweep
CPMM
Friday close is held. Monday opens –8%. Arb bots clear the pool in seconds; LPs realize the entire gap as loss.
Oracle-aware
Factor-adjusted band tracks ES futures across the weekend. AMM quote is at fair value Monday open — no free arb to capture.
03
Vol regime spike
CPMM
Fee tier is static. Vol jumps from 12 to 35; LPs are paid the same to hold a 3× riskier book.
Oracle-aware
Regime label switches the served forecaster (F1 → F0 in high_vol). Band widens; AMM's band-conditioned execution adapts in step.
04
Halt + multiplier lag
CPMM
Token-2022 multiplier updates with delay during halts / corporate actions. Trades execute in the inconsistency window at stale prices.
Oracle-aware
Regime = Halt. Band CI widens; AMM gates trading or quotes only inside-band, with the receipt declaring why.
05
Primary-oracle drift
CPMM
Primary feed has an incident or drifts off fundamentals. Pool keeps trading on stale data; LPs eat MEV / toxic flow until somebody notices.
Oracle-aware
Soothsayer's calibrated band is an independent cross-check. AMM falls back to the band with explicit uncertainty disclosed in the receipt.
Every scenario: the AMM has to price something. Soothsayer makes that price uncertainty-aware — and publishes the receipt.
Soothsayer.
08 / 12

Wedge, TAM, outcome.

2026 → 2030 · AMM lens
First-pair proof
[TBD]k
annualized fee + LVR recapture · first xStock pool
Single-pair AMM MVP sized to validate band-conditioned execution. The same architecture scales to every xStock and to the future stack (perp, options, settlement) above.
2030 outcome
$150M–$1.5B
valuation range
Comparable multiples — spot AMM: Lifinity, Hashflow, Orca on 2030 ARR. Oracle-feed: Pyth 20–40×, RedStone 15–55×. Soothsayer captures both.
L1 · TAM
Tokenized RWA, global
$9–16T
L2
Tokenized equity subset
$200B
L3
Solana share @ 50%
$100B
L4
xStock annualized spot DEX volume
[TBD]
L5 · SAM
Oracle-aware AMM addressable (fees + LVR recapture)
[TBD]
L6 · ARR
Soothsayer AMM-share 2030
[TBD]
Soothsayer.
09 / 12

Business model.

Revenue vectors · ranked
01 · PRIMARY
AMM fees + LVR recapture
Trading fees on the oracle-aware AMM, plus the LVR that band-conditioned execution captures back from CPMM LPs. Routed protocol-fee share funds treasury and (future) token holders.
Devnet → mainnet 2026
02 · SECONDARY
Band-feed licensing
Per-read / per-subscription access to the calibrated band — for other AMMs, perp DEXes, structured products, and lending markets. Pyth-style B2B distribution, validated by the AMM running on top of it first.
After Papers 1 + 3
03 · TERTIARY
Design-partner risk work
Reserve-buffer analysis, liquidation-policy studies, and closed-market risk reviews built on the Paper 3 evaluation stack. Lending protocols (Kamino, MarginFi) and risk consultancies as design partners.
After Paper 3
04 · LONG TERM
Future-stack primitives (perp, options, settlement)
Each downstream primitive uses the same band and the same receipt — perp mark/liquidation, calibrated options vol, licensed settlement to third-party venues. Independent product lines on shared infrastructure.
2027+
Operating cost $310–800 / mo  — the moat is calibration, not infra.
Soothsayer.
10 / 12

The humility slide · push back on any of these

Risks and honest unknowns.

01
Incumbents could close the calibration gap. If Pyth or RedStone ship verifiable realized-coverage receipts, the band's moat narrows from methodology to integration depth. We monitor; the AMM's defense becomes the on-chain mechanism, not just the band.
02
Single-team credibility. Pyth runs 100+ institutional publishers. Soothsayer ships one model and one team. Multi-publisher trust has to come through replication of the methodology, not BD.
03
8-day AMM execution risk. Colosseum target is a devnet MVP on a single xStock pair, not a mainnet-ready protocol. Investors expecting production-grade audited code at submission will see a working demo with disclosed gaps.
04
LP / liquidity bootstrap. The oracle-aware AMM is most defensible with meaningful LP capital and routed flow. The devnet demo needs seeded liquidity + simulated arb pressure to show LVR-recapture economics live.
05
Incumbent AMMs could bolt on band-style oracles. Lifinity, Hashflow, and concentrated-liquidity DEXes could plug a Pyth-confidence dial into fee curves. The defense is depth — calibrated band + receipts + xStock-specific factor switchboard, not a single-axis oracle plug-in.
06
Distribution is concentrated in incumbent stacks. xStocks Alliance, Kamino, Ondo, and Backed already sit on established feeds and DEX surfaces. We have to win by making the calibration + AMM stack the obvious upgrade, not by replacing distribution rails.
Soothsayer.
11 / 12

The ask.

Roadmap · edge · close
Phase 0 ✅
complete · 2026-04-25
Band validation
12-year decade-scale backtest, 5,986 panel rows across 10 tickers. Full PASS — Kupiec + Christoffersen, OOS 2023+, at τ ∈ {0.68, 0.85, 0.95}. Rust port at byte-for-byte parity with Python (75/75 tests).
Phase 1
now · 8 days · Colosseum
Devnet AMM MVP + video
Oracle-aware AMM on a single xStock pair, devnet. Video submission for Colosseum. Paper 1 §4/§5/§7/§8 → arXiv (q-fin.RM, ACM AFT) in parallel.
Phase 2
post-Colosseum · ~6 weeks
Mainnet · papers live · design partners
Mainnet AMM on first pair · public comparator dashboard (Soothsayer vs Chainlink vs Pyth) · Paper 1 + Paper 3 on arXiv + SSRN · first design-partner conversations (Solana DEX teams, Backed, Ondo, Kamino BD).
Phase 3
gated
Multi-pair · B2B band feed · future stack
Gated on AMM TVL + Paper 3 evidence + ≥1 design-partner LOI. Multi-pair AMM expansion · band-feed B2B licensing · future-stack specs (perp, options, settlement).
Specific asks
A Feedback on positioning — Soothsayer Research + the oracle-aware AMM as a single product stack.
B Intros to Solana DEX teams + market makers + RWA issuers (Lifinity, Orca, Meteora, Jupiter; Wintermute / Selini / Flowdesk; Backed, Ondo, Kamino BD).
C Superteam-brokered Solana RPC for historical pool-state and Kamino tx backfill.
My edge
ML engineer × Rust × DeFi protocol design.
Conformal-prediction calibration discipline (held-out, ablation, bootstrap CIs). Rust + Anchor in the language Solana runs in. Protocol-design fluency — LVR-aware AMMs, oracle-conditioned pricing, intent execution.
The overlap of these three pillars is vanishingly small in the Colosseum submission pool.
Soothsayer.
12 / 12