Solana is about to replace its entire consensus. Finality drops from 12.8 seconds to ~150 milliseconds. I have been testing Alpenglow on the community cluster since its first genesis - so I wrote the longread I wish existed: what changes, why it matters, and how it looks from the inside. cryptovik.info/news/solana-a…

Aug 7, 2026 Β· 2:52 PM UTC

20
42
209
129,145
The cluster runs on people: @Ashwinningg and @TimGarcia0 at the core, @solblaze_org distributing the cluster's stake and backing volunteers from his mainnet pool, @Allnodes with free servers, @PuffinValidator inventing new metrics - and dozens of operators on their own hardware.
1
12
3,520
Sort replies: Relevant Recent Liked
Replying to @hvzp3
Excellent stuff
1
2
369
Replying to @hvzp3
πŸ™πŸΌπŸ™πŸΌ thanks!
1
1
148
Thank you, Simon! 🀝
97
Replying to @hvzp3
Nice read
1
1
189
Replying to @hvzp3
Great job!
1
1
151
Replying to @hvzp3
Great job πŸ‘
1
1
499
Thanks! πŸ”₯
1
407
Replying to @hvzp3
It was a great pleasure to read it! Hats off to you: everything is clear, explained, and there is inspiration on top of that!πŸ”₯πŸ™
1
4
800
Replying to @hvzp3
🧠🧠🧠 Great job! πŸ‘
1
2
227
Thanks, Cherry! The support of our Ukrainian Solana community keeps me inspired to do more work like this πŸ”₯
182
Replying to @hvzp3
TowerBFT + PoH out Votor in ~150ms finality not a patch a consensus transplant great operator-level write-up, colleague ⚑️
1
1
458
Thank you, colleague!🀝 and let's continue validating!πŸ’»
1
331
Replying to @hvzp3
Any jepsen tests? Those will be crucial for simulating real network failures.
1
209
Not that I have seen so far. Beyond the formal proofs, audits, and the current bug bounty, Alpenglow still has full testnet and devnet cycles ahead of mainnet. This is usually where failure testing takes place. A Jepsen-style suite would be a nice addition though.
1
202
Replying to @hvzp3
Run it the fuck up
1
178
Replying to @hvzp3
Best work, thx for your time!
1
36
Replying to @hvzp3
Well done πŸ™
1
39
Replying to @hvzp3
finality drops from 12.8 seconds to ~150 milliseconds – not a typo? thatβ€˜s gargantuan πŸ‘€
1
197
Replying to @hvzp3
Great job!
1
47
Replying to @hvzp3
πŸ”₯
9
Replying to @hvzp3
we'll see how many validators survive the cut.
206