Follow up to my earlier post questioning the CertiK TPS verification. I've spent the past week actually reading the architecture documentation, the CertiK report, and the consensus mechanism papers. Here is my assessment of what checks out and what remains uncertain.
What I think holds up:
- The bare metal, no VM architecture explanation for the TPS figure is technically coherent. EVM overhead is real and eliminating it does allow substantially higher execution throughput. The number is still extraordinary, but the mechanism is sound.
- The 451/676 quorum is formally Byzantine fault tolerant at the minimum threshold. The math is correct: 676 >= 3(225)+1 = 676. Whether minimum viable BFT is the right security model is a different question, but the formal claim is accurate.
- tick-based consensus (replacing block propagation delays) also checks out as a legitimate architectural decision that removes a meaningful throughput bottleneck.
What remains uncertain:
- The useful PoW claim that mining compute contributes to AI training while maintaining security properties has not been independently verified in the same way the TPS figure has.
- The 676 node validator set is small. BFT at this scale has different attack surface characteristics than e.g. Ethereum's validator set.
- The DOGE mining integration they're teasing hasn't launched yet, so any claims about its scale are forward-looking.
My overall read after a week of digging the architecture is more technically interesting than I expected going in. I remain skeptical of the promotional framing around some of the numbers, but skeptical doesn't mean to say wrong.
The onchain hashrate data for DOGE mining will be interesting to see!
