We separated church and state because a king who spoke for god never had to ask.
Now let's separate money and state.
The printer funds wars nobody wants.
It buys elections with money that doesn't exist.
It taxes you forever through inflation.
A state that can't print money must ask.
How Flint works
The book lives on-chain. Matching and settlement both happen in the Solana program; no off-chain engine holding the real state. The server is an edge, not a sequencer: landing, market data streams, analytics.
@thedavidgorski on the full system 👇
Many firms have pricing models. ~Five have the Solana engineering.
Right now, it's not easy or cheap to provide active liquidity on this chain.
I want a big, beautiful, ergonomic on-ramp for new blood and that's what I'm focused on.
Soon.
worth mentioning re this Solana PERP DEX: Superis Labs project:
- this is already well advanced
- there are some very smart people working on this
- this will be good
Solana Transaction Execution Lock Relaxation Proposal
A scheduler (both leader and replay) change that has a theoretical improvement of less than ~2x. No protocol changes needed.
Idea: Drop transaction read locks right away after loading the account state.
Why? This allows upcoming transactions that write-lock the same account to start executing immediately. Pipeline efficiency potential increases significantly (depending on shape).
How? Enforce linearizability requirements. While sub-sequent transactions write locking the same account can start execution, they cannot be recorded as before the reading transaction.
All account states are locked, read and frozen at the start of the transaction; this keeps things consistent throughout execution of the reading transaction.
The linearizability concept happens at the commit/record level; even though the reading and writing transactions might be happening at the same time; the read will always be recorded before the write.
In the case of failure; write transaction is unblocked from commit.
Now that transaction batches can be conflicting (recent change), the trade off here would be the savings of locking and loading the account in one thread and then execute sequentially versus finding an optimized boundary in the graph to split and put into another thread (paying the price for the cross-thread communication). For 'heavy' transactions this could exceed the cross-core optimization.
Generally the less restrictions for ordering, the more flexibility the execution pipeline has and the more efficiently it can schedule.
Flow of Value > Store of Value
Technologies, systems and protocols that enable compute, dynamic state changes, and real transactions are far more interesting than static, HODL til you die, sinks for capital.
Holding has it's place; but it is inherently stagnating on it's own.