Benjamin Matsumoto
@zkJyu
building and growing @zkPass, the verifiable internet for AI agents and humans | prev @UOsaka_ja @SoftBank
1.1K Following 2.3K Followers
Giving an agent a wallet does not make it a credible counterparty. Neither does an identity or a benchmark score.
Credibility comes from evidence that survives scrutiny and accumulates over time.
is building the infrastructure for that evidence to travel with the agent, across auditors, applications and markets.
Show more
Soon your agent will hire other agents, spend your money, and make decisions you never see.
Right now, it has almost no way to know which of them are actually good at what they claim. That should make everyone a little uncomfortable. This is what has kept us busy at @zkPass.
Show more
Every major technology wave creates the same problem: trust.
The internet solved information exchange, but needed search to find what matters. Blockchains solved ownership, but needed settlement to move value.
AI agents will solve execution, but they will need a new layer to determine what can be trusted.
As agents begin to interact, transact, and operate autonomously, verification becomes the foundation of the economy.
Show more
There’s a huge difference between "internet data" and "verifiable internet data".
Most of the difficulty is in the connection layer.
Most AI systems today are built on an implicit assumption
that the data they consume is "good enough"
In reality, they operate on a mix of scraped, cached, proxied, and often unverifiable inputs
That works for content generation
It completely breaks down for anything involving capital, identity, or decision-making
The moment AI moves from generating text to taking actions
data integrity becomes the bottleneck
Verifiable data is not a nice-to-have layer
It is the prerequisite for turning AI systems into accountable systems
Show more
