Patrick O'Grady & Brendan Chou | The Coordinate Podcast with Soubhik Deb скачать в хорошем качестве

Patrick O'Grady & Brendan Chou | The Coordinate Podcast with Soubhik Deb

In Episode 002 of TheCoordinate, Soubhik sat down with Patrick O’Grady and Brendan Chou from Commonware to answer a deceptively simple question: What does state-of-the-art blockchain engineering look like in 2026? Their answer isn’t “pick the right framework.” It’s: stop inheriting monolithic stacks, and instead compose a chain from optimized primitives you can swap, tune, and evolve safely over years. What we cover The “Anti-Framework” thesis: specialization by construction, not endless framework forks Composable primitives: runtime, consensus, networking, crypto, storage, broadcast, and more—built to be independently swappable Deterministic distributed testing: reproduce real distributed failures with seeded scheduling/timing (goodbye ghost bugs) Consensus trade-offs (Simplex → Minimit): assumptions, failure modes, complexity budgets, and why “small” theory changes can explode operational risk The scaling reality: fast consensus creates a data firehose, broadcast, storage, indexing, and client access become the real bottleneck Open + commercial split: Commonware (open) as the base stack, ExoWare (commercial) as the “around-the-chain” infra layer (data services, indexing, low-latency access, tooling) Who this is for If you’re building a chain/appchain/L2 and feel constrained by standard frameworks—or you care about high throughput + low latency, deterministic testing, and engineering-first consensus discussions, this one’s for you. Subscribe for more deep technical conversations on modern systems, crypto infrastructure, and pragmatic engineering trade-offs. 0:00 Show mission + episode setup (specialization vs monoliths) 1:18 Meet the guests: Patrick O’Grady + Brendan Chou (Commonware) 1:30 The thesis: “anti-framework” primitives for specialization 2:26 What “broke” Patrick: forks, abstraction debt, fear of changing anything 4:26 Bitcoin → OS/framework era → why the model is starting to hinder teams 6:34 Commonware’s approach: decomposing the stack into 40+ primitives 11:10 The big primitives: Simplex + runtime + “knobs to turn” 13:03 Deterministic simulation + fuzzing for distributed-systems testing 16:59 Authenticated validator-only P2P (no TLS cert complexity) 18:33 Serving data to non-validators: relays/DA choices left to the app 19:49 Cryptography layer: threshold BLS, safer interfaces, parallelism + math ops 23:09 Framework tradeoffs: faster launch vs painful deep customization (Cosmos/L2s) 25:31 Brendan on dYdX/Cosmos: where the framework boundary starts to bite 28:38 Minimit: why a 20% Byzantine model enables much faster consensus 30:51 Alpenglow inspiration: two-phase finality + “one-shot” finalization idea 37:49 The Minimit twist: faster view changes (40% notarization threshold) 41:46 Using Commonware primitives to build/benchmark Minimit end-to-end 42:59 Throughput reality check: broadcast, storage, and the “tons of data” problem 46:31 ExoWare: infra around validators (querying/streaming/relays) without holding keys 52:53 What’s next: Minimit, async DKG, threshold encryption, optimized state DB 56:42 Wrap-up + where to find Commonware