Requant Lab

Requant Lab is the network side of Requant's research. The research programme settled what one unit of work is: TNet v1, checked by one row. A coin is also a network: nodes that hear about blocks late, pools that split, difficulty that has to follow hashrate, miners who can cheat. The lab plays such networks in software, so these questions can be asked before they are met on a live chain.

What it is

• A laboratory of virtual proof-of-work networks. A deterministic event engine in Rust plays nodes, pools, miner groups, links, blocks, transactions and difficulty rules over virtual time. One seed gives one history; the same build, scenario and seed give the same history again, checked by hashes.
• A console in the browser at requant.network/lab: open a world, play it back, build an experiment on a timeline, take the run apart in tabs, compare it with the same world without the experiment. See Using Requant Lab.
• Live broadcasts at requant.network/lab/?channel=bitcoin: a network that plays by the server's clock and that viewers steer together. See Live broadcasts.
• Real Requant blocks. A protocol mode replaces the model's blocks with real Requant regtest blocks, each checked by every simulated node with Requant's own consensus code. See Real Requant blocks in the lab.
• A server and an API for runs, series over many seeds, long-lived worlds and stories for the map. See Requant Lab API.

Why Requant needs it

• The test network is small. Since 9 October 2026 the Requant testnet has run on a few seed nodes and a handful of miners, and blocks have been intermittent. Network-level effects (deep reorganisations, partitions, a dominant pool, a difficulty rule under a hashrate swing) cannot be waited for there, and must not be provoked there.
• Rules can be compared before they are chosen. The lab plays the same network under different difficulty rules, propagation models or pool strategies on the same seeds, and reports the difference with intervals over many runs instead of one anecdote.
• The model is held to the real code. The model's ASERT difficulty matches Requant's consensus bit for bit on whole seconds, and the protocol mode runs the real block checks. Where the model simplifies, the coverage table says so.

How it is built

• One engine. Since 30 October 2026 everything runs on one fast engine, the protocol mode included. Its exact mode writes the same event journal, byte for byte, as the step-by-step engine it replaced; that engine's event loop is removed, and frozen reference values keep the comparison.
• The same code everywhere. The browser runs the engine as WebAssembly in a Web Worker; the server runs the same Rust natively. A story built on the server opens in the browser at any moment without replaying it.
• Independent of the coin's code. The core engine has no dependency on Requant; the Requant adapter is a separate workspace pinned to a Requant commit.
• Measured, with conditions. On 17 October 2026, on one core of an AMD EPYC 9654 server under a one-core service limit, the engine of that date processed 125-290 thousand events per second; a year of virtual time took 9 s for 2 nodes and about 2 minutes for 8 nodes, and the browser was about 8 times slower. The protocol mode takes 13-20 ms per real block.

What it does not claim

• A simulation, not a measurement. Every number the lab produces is the output of a model under the stated scenario and seed. It is not a measurement of the Requant testnet or of any real network.
• Real networks are scaled. The profiles of Bitcoin, Litecoin, Dogecoin, Bitcoin Cash, Zcash, Dash and a Monero-like network keep their block interval, difficulty rule and emission schedule, but emission runs 100 times faster, transaction flow and block size are 20 times smaller, and the network is 60 nodes on a world map.
• Bots are scripted. The viewers the broadcast server keeps follow fixed probabilities; they show what a crowd can do to a network, not what people will do.
• The source is not public yet. The lab's repository is private for now; the lab, its API and its contract are public.

Where it is going

The lab's milestones are on the roadmap: attacks of the fee era (undercutting, stubborn and optimal selfish mining, bribery), proof-of-stake networks, votes and predictions in the broadcasts, and a broadcast of your own plan.