nobody wants confidential compute. they want to stop worrying about getting breached.
my pitch used to be “verifiable confidential compute platform.”
now it’s “we host software you can’t afford to get hacked.”
both true. yet, completely different conversations.
Early Caution customers are choosing bring-your-own-compute.
The pattern is showing up in inbound too: teams want verifiable compute to run where their infrastructure already lives.
We didn’t want an audit just for the sake of saying we had one, like many crypto projects do.
Too often, projects pay an audit firm for the optics of having their code "greenlit" for investors, only for major vulnerabilities to still exist beneath the surface. In many cases, it feels more like checking a box to make users feel comfortable rather than genuinely hardening systems, something we've seen repeatedly across crypto, including incidents like Tenexium Subnet 67 within the $TAO eco.
We wanted a team that would actually add value.
Distrust stood out from the very first conversation. Their depth of knowledge went far beyond our expectations, and they immediately identified ways we could strengthen our systems before we had even formally engaged them. I'm genuinely glad we chose to work with them.
They stayed highly engaged throughout the review process and helped us focus on the risks that actually mattered while we hardened our systems in parallel alongside their audit work and reporting.
There are some very exciting things ahead that we think users are going to appreciate. TrustedStake will be among the first platforms in the world to introduce a new standard for verifiable execution and infrastructure security.
More on that soon.
Big thank you to the entire Distrust team.
TEEs are a more secure way to run software and the biggest obstacles to their adoption have always been deployment complexity and verification workflows.
Once those become easy, TEEs are a no-brainer for a huge range of applications, especially blockchain co-processors.
The recent ZeroLayer incident is a good example. Code was modified and there was no simple way to independently verify what was actually running on the nodes that mission critical systems depended on.
That’s why I spend so much energy making this technology easier to use. It should be as simple to deploy and operate as any other server hardware.
Everything that matters should be verifiable in real time.
Had you told me a year and a half ago that I'd be part of the founding team of a confidential compute platform, I wouldn't have believed you.
And yet, Caution is live in closed beta 🥹
Six months from first commit to here.
@antonlivaja, Lance, Ryan and I have built something I'm genuinely proud of, and now we're putting it in front of our first customers in production.
Our confidential compute platform is live in closed beta 🎉 We're working hands-on with the first customers, helping teams migrate existing TEE workloads or stand up confidential compute for the first time.
Translation: even teams already using secure enclaves mostly can’t turn attestation into real proof of what code is running.
They have a PCR hash. They can’t reliably reproduce it from source or validate the full chain of trust. They’re trusting a number they can’t independently verify.
That’s not a security property. That’s a black box with better marketing.
This is exactly what we’re building @CautionHosting to fix.
84% of orgs say attestation validation is a top barrier to confidential computing adoption (vs. just 77% for “niche tech” perception, 75% for skills gap).
CCC x IDC survey of 600+ IT leaders, Dec 2025:
https://t.co/8EAmhCS41O
"I want my AI agents to be able to do everything for me, so I gave them root access, but I want my setup to be as secure as if I didn't give them any permissions or access." - everyone
we built an open source verifiable compute framework to submit as part of the zypherpunk hackathon. check it out (there is a cool demo video): https://t.co/57K2VJCFyQ
As part of the demo we ran a zebra node in a fully verifiable manner, as well as an llm.
if you want to skip straight to the code here you go: https://t.co/zpsWFiQtJC
we built an open source verifiable compute framework to submit as part of the zypherpunk hackathon. check it out (there is a cool demo video): https://t.co/57K2VJCFyQ
As part of the demo we ran a zebra node in a fully verifiable manner, as well as an llm.
if you want to skip straight to the code here you go: https://t.co/zpsWFiQtJC