Pinned Tweet
Acrux Net, 33 seconds.
A Solana fork with 50 ms slots. Three hubs, FRA, NYC and TYO, and every slot built by three validators that seal their blocks until all three commit, then merge into one order. 324 ms to finality.
It's running right now: acrux.computer/net
Ran stock web3.js onSlotChange against Acrux and Solana mainnet at the same time, from one client, for 120 s.
Median slot gap: 50.1 ms on Acrux, 201 ms on Solana.
Replayed 2,012 real Solana mainnet slots through Acrux's merge.
48 sandwiches, 66 people caught in them. Sent 2-of-3 on Acrux, the 54 we could price would've kept 0.153 SOL of the 0.192 the traders took.
acrux.computer/replay
teapawt.org/AcruxComputer/status/2…
Replying to @ox_legendaryy
yep, Backpack with a custom RPC. next up is replaying real mainnet slots through Acrux's merge, so an app can see what its users would've kept from sandwiches before it moves
one of those slots, 470,916,233. a trader bought at 2, two wallets bought behind it, the trader sold at 340
sent 2-of-3 on Acrux there's no front leg, and the first buyer gets 23,551,969 tokens instead of 22,755,405
acrux.computer/replay/470916…
Anyone can follow Acrux Net from their own machine now.
acrux node follow, pointed at rpc.acrux.computer: it re-executes the live chain slot by slot and every state hash matches. Works since node 0.4.1, source is up:
github.com/AcruxComputer/acr…
Co-location sounded like the hard sell when we started.
Then we read the paper's table: with a 25 ms local delay and a 1% minimum hub stake, only 1.2% of Solana's stake sits outside a city that could already form a hub.
Fault Lab, censorship: two of the three proposers leave a transaction out on purpose.
The third one carries it, its block is in the agreed set, and the transaction lands in slot 66 anyway. Keeping it out takes all three.
acrux.computer/lab/2916f539-…
Anchor works on Acrux now.
Took a counter program, set the provider cluster to rpc.acrux.computer and changed nothing else. anchor deploy, then anchor test: 3 passing.
And since 0.4.0 the chain keeps its state through node updates, so what you deploy stays.
Fault Lab, last look: one proposer holds its commit back to see the other two blocks first.
It comes in 28 ms past the 20 ms window, both co-proposers mark it late, its block gets dropped and the slots skip. Lemma 2.3, in a sandbox anyone can start.
acrux.computer/lab
The stock solana CLI works against Acrux now. Only the URL changes.
Sent a transfer, solana confirm -v came back Status Ok, 150 CU, fee 0.000008 ACRUX.
acrux.computer/docs/sdk/sola…
web3.js too. sendAndConfirmTransaction confirms on Acrux over the same subscriptions it uses on Solana, found from the RPC URL alone.
From about 290 ms away it took 2.9 s. Next to the net, 0.3.
solana program deploy works on Acrux too.
Deployed a small program, 45,928 bytes, and called it. It logged "called with gm, Acrux", 3,140 CU.
Any Solana program goes on the same way: acrux.computer/docs/sdk/sola…
Acrux Net, connected and used through Backpack.
Most wallets label the native gas of an SVM chain as SOL, so that's what it shows.
The transfer went over Acrux.
acrux.computer/docs/start/ad…
Gas and stake on Acrux. It comes over a bridge from Solana, then at one epoch the chain switches its fees and stake to it.
Today the net runs on test ACRUX from the faucet.
Launch traffic stopped sends on Acrux twice today, 17:54 to 18:01 and 18:04 to 18:52 UTC. ~75 people on the live event stream, and the validator behind the gateway fell behind.
Running again since 18:52: 9,369 slots, 0 skipped cblocks by 19:01. The live stream on /net stays paused until the gateway moves off the validators' runtime.









