The 22-Hour Timeline

The Most Important Competitive Signal

On March 31, 2026, Alex Kahn posted a thread revealing Dropset. His new onchain CLOB on Solana. Twenty-two hours later, on April 1, Sital Kedia shipped Bulk Orders on Decibel.

The 22-hour gap is not a coincidence. It is proof of three things simultaneously:

1. 02Alex identified the correct problem and the correct solution
2. 04Sital recognized it immediately. The fastest possible engineering acknowledgment
3. 06Alex is leading this design space. Decibel is reacting

The Problem Both Teams Were Solving

99% of market maker orders are cancelled before they match. Yet market makers pay full gas to index all of them in the orderbook. On every CLOB. Onchain and off-chain. The dominant activity is order management, not order execution.

The specific cost breakdown for a Solana or Aptos market maker:

• ›Submit a ladder of 20 quotes: pay gas on 20 transactions
• ›Market moves: cancel all 20. Pay gas on 20 more transactions
• ›Repost the ladder at new prices: pay gas on 20 more transactions
• ›Net result: 60 transactions, 0 trades executed

At any meaningful HFT frequency, this is economically unviable. The gas cost of maintaining a competitive two-sided market exceeds the spread revenue from the occasional fill.

Both Alex and Sital independently built the same solution to this problem: atomic batch ladder replace. One transaction cancels the old ladder and places the new one. 60 transactions collapses to 1.

Alex's Dropset

Revealed March 31, 2026.

The core primitive is drop-and-set:

• ›Drop phase: atomically cancel the entire order ladder in one operation. Silent return if the drop fails, no race condition between cancel and replace.
• ›Set phase: atomically place the new ladder. If drop succeeded, set proceeds. If drop was silently skipped (no orders to cancel), set places fresh.
• ›Combined: one atomic transaction that achieves what previously required N cancel transactions followed by N place transactions.

Dropset's engineering stack is extraordinary given Solana's hostile runtime:

SBPF assembly with RB-trees. The matching engine is written directly in Solana BPF assembly, using red-black tree data structures for O(log n) worst-case orderbook operations. Assembly gives direct memory layout control that eliminates the overhead of high-level language abstractions.

Single canonical PDA with one ATA vault. All orders for a market live in a single program-derived address with one associated token account. Eliminates the account proliferation that makes Solana programs expensive. Fewer accounts to declare upfront means lower transaction fees.

Absolute pointer addressing with pinned empty signer. A specific Solana optimization: by pinning an empty signer account at a fixed memory address, pointer resolution for the signing operation becomes a single cycle rather than a table lookup. Microsecond improvements at HFT scale.

Chunked 64-bit pubkey compares. Comparing 32-byte Solana public keys normally requires 4 sequential 64-bit comparisons with branch prediction uncertainty. Chunked compare parallelizes these comparisons using SBPF's specific register file layout, reducing the comparison to near-single-instruction latency.

CLRS spec compiled to Lean formal proofs. The matching engine's correctness properties. Price-time priority, no double-fills, no money creation. Are specified in CLRS pseudocode and compiled to Lean 4 theorems. Formal verification of a CLOB is unprecedented. This is the first CLOB in existence with machine-checked correctness proofs.

Status as of April 2026: 100% open source (GitHub public repository), pre-mainnet beta, multi-assembly refactor in progress. Alex has not yet launched commercially on Solana.

Sital's Decibel Bulk Orders

Shipped April 1, 2026. Less than 22 hours after Alex's thread.

The implementation:

public entry fun place_bulk_orders_to_subaccount(
    trader: &signer,
    market_id: u64,
    // New desired order state
    orders: vector<OrderSpec>,
    // Sequence number prevents out-of-order execution
    sequence_number: u64,
    // Compact vector for non-top-of-book levels
    // reduces serialization overhead
    compact_levels: vector<u8>,
) {
    // Atomically cancel all existing orders
    // Replace with new order set
    // Single BigOrderedMap B+tree operation
}

Key properties:

• ›Atomically cancel + replace in one transaction (the core Dropset insight)
• ›Compact vector encoding for non-top-of-book orders (reduces transaction size)
• ›Sequence numbers prevent out-of-order execution when multiple Bulk Order transactions are in flight
• ›Partial failure tolerant: if one order spec fails validation, the rest execute

The announcement did not mention Dropset. No reference to Alex's thread. No acknowledgment of the prior art. But the 22-hour response time makes the competitive dynamic unambiguous to anyone who tracks both projects.

Decibel's version has one acknowledged advantage: it's written in high-level Move 2, not SBPF assembly. Move's type safety and resource model make the code safer and more auditable. Solana's hostile serialization constraint forced Alex to fight at the assembly level. Aptos's more accommodating runtime lets Sital write clean, safe code.

The Technical Comparison

The core technical verdict:

• ›Dropset is more impressive engineering. Alex fought a hostile runtime with assembly-level optimization and formal verification
• ›Decibel Bulk Orders is cleaner. The Move 2 implementation is more readable and safer, and Aptos's runtime makes it possible to write at a higher level
• ›Both solve the identical problem with the identical atomic primitive

Why the 22-hour Response Matters

It proves Alex was right about the primitive. The fastest possible validator of a technical insight is a competing engineer independently implementing the same solution. Sital shipped Bulk Orders in 22 hours because he saw immediately that atomic cancel-and-replace was the correct primitive. That kind of response only happens when you see something that's obviously right.

The reaction is specifically to the primitive, not to Decibel as a product. The Decibel team has been building the perps engine since mid-2025 and it has been live on mainnet since February 2026. Decibel isn't a 22-hour project. What happened in 22 hours was that an engineer on a production exchange saw a better way to handle quote laddering and slotted it in. The acknowledgment is narrow and technical: this specific primitive is the right shape.

It validates the architecture on Aptos. The question "will atomic ladder replace actually work in production on Aptos?" is answered by Decibel's own behavior. They shipped the primitive, on Aptos, within 22 hours of the concept becoming visible. It works in production.

It sharpens the opportunity for Whop. Econia was the first fully onchain orderbook on Aptos, built by Alex, and later shut down. Alex left Aptos for Solana to build Dropset. Decibel now covers perps on Aptos. Spot is still open. The person who invented the fully onchain atomic spot CLOB is available, and the window to bring him back to Aptos for Whop's spot product closes the moment Dropset gets commercial traction on Solana.

Whop's Spot CLOB Advantage

Neither Dropset nor Decibel Bulk Orders have UTT integration. Position sizes are visible in both systems.

Whop's spot CLOB adds:

• ›Encrypted mempool hides orders before they hit the book. MEV protection at the pre-execution layer. Settlement uses CA/UTT for post-execution transfer privacy
• ›The taker sweep is atomic and crankless (Decibel perps uses a permissionless crank. Sandwich attack vector)
• ›Per-user seat tracking makes cancel-all O(n user orders), not O(book size)
• ›Formal verification (Alex's CLRS → Lean pipeline from Dropset, ported to Move)
• ›Chart replay auto-triggered on every fill
• ›Content Rewards attribution on every trade

The Dropset insight is the foundation. The UTT integration, crankless design, and attribution infrastructure are the Whop additions. Alex builds the foundation. The Whop stack adds the privacy and social layer on top.

Key Dates