Bridge Atlas - Episode 5: Verifiability | Shreya Shankar & Justin Drake
Shreya Shankar (Berkeley PhD, creator of the DocETL declarative data-processing system) and Justin Drake (Ethereum Foundation, lean Ethereum project) compare verifiability across AI data pipelines and blockchain consensus. Drake explains SNARKs, trust-minimization, and recursive proof composition for Ethereum's deterministic and non-deterministic components, while Shankar details how DocETL handles LLM-powered query pipelines, human-AI misalignment, and eval design for unstructured data like court transcripts, with both drawing parallels between 'hints' in SNARK proving and decomposition techniques for AI reasoning verification.
Shreya Shankar, Justin Drake
