Writing from Provably.

In 9 of 24 repaired runs, no model was asked to write a new answer after recovery. Where synthesis did run, an execution ledger raised supported values reaching the answer from 73.9% to 81.2%, a direction but not yet a significant effect.

On the same 60 failed runs, SourceryKit-guided recovery restored complete tool evidence in 53.3%, compared with 30.0% for blind retry. But repairing the evidence did not always repair the final answer.

We benchmarked 15 leading AI models and found SourceryKit detected 100% of the covered deterministic errors—including 109 unsupported claims that answer-only scoring accepted.

Introducing runtime verification for agent tool calling: how SourceryKit lets production agents prove and verify tool outputs at runtime, catch errors before they propagate, and move closer to reliable autonomy.

Provably V2 is now live and available for anyone to access and use. Allows users to connect to Relation DBs, comes with an SQL IDE to run verifiable queries at speed.

Why QEDB makes verifiable SQL practical, without circuits or heavy SNARK overhead.

QEDB is Provably’s take on verifiable databases that ditches “SNARK everything” thinking and goes back to first principles. The post explains why trusting remote database answers is still a mess in modern stacks, surveys the two dominant VDB lineages (authenticated data structures vs SNARK-based systems), and then introduces QEDB as a third path.

Major teams are building confidential AI infrastructure to store and process sensitive workloads.

Provably and Midnight are collaborating to advance the next generation of ZK on big data. The partnership combines Provably’s verifiable analytics with Midnight’s privacy-focused blockchain infrastructure to enable confidential DeFi, enterprise-grade privacy, and developer-friendly ZK tooling.

Provably’s new protocol, QEDB, introduces a faster, simpler way to verify database queries without SNARKs. Built on specialized cryptographic data structures, QEDB supports complex SQL queries, scales to millions of rows, and produces constant-size proofs under 1 KB.

We are entering a golden age of AI agents, where they bring tremendous convenience to many aspects of our workflows and lives.

Provably V1.3 is live! This release marks the debut of the Verifiable Analytics Network, connecting Data Owners with Insight Seekers as well as expanded SQL aggregation support.

This is a unique blog that I co-authored with Emanuele. We first met at a ZKWarsaw meetup, where I gave a presentation...

Provably V1 simplifies the complex world of Zero-Knowledge (ZK) proofs, empowering you to manage access to your data with confidence, knowing it remains private.

Find a simple way to give proofs that information and analytics insights are created by aggregated queries on data stored in private relational databases.