Everyone’s calling him The Privacy Guy now...
We’ve been calling him that since @GuyZys was teaching devs about onchain privacy - before it was the next 1000x narrative.
x402 is really cool but it's missing one critical component: privacy.
Really fun experimentation by our engineer @rogue_rotkosky merging FHE and x402.
Check out the app! https://t.co/3M31UxI6zX
Cool proposal from Carter.
Encrypted intents powered by FHE unlock a fairer, MEV-resistant coordination layer - something we’ve been building toward with CoFHE.
When computation stays private and verifiable, everyone wins.
"If your privacy depends on a vendor’s attestation key, you have a service agreement - not privacy."
Watch the full recording of TEE vs FHE discussion for more hot takes:
https://t.co/UloD7b4mVw
Three things just changed:
1/ FHE decryption got 20,000× faster
2/ Private DeFi is unlocked
3/ Fhenix made it happen
READ HOW ↴
https://t.co/8fZbB5JJPN
TEEs had a good run, but this @intel response is exactly why we've been building toward FHE-first systems.
The attack itself? Not shocking. But Intel basically saying "not our problem" kills any enterprise adoption path for TEE-only privacy.
Here's what people miss: this isn't just about SGX dying. It's about what comes next. Pure MPC is still a non-starter for most applications - too slow, too complex, breaks with scale. But FHE? That's your new building block for the web.
With FHE, your compute happens on encrypted data. No trusted hardware needed. No "trust the box" assumptions. Just math protecting your private state while still letting you build programmable applications on top.
The era of "good enough" privacy is over. Time to build the real thing.
The latest in TEE drama: I’ve known about WireTap for a while; paired with Battering RAM, it reinforces a point many of us have lived with for years: TEEs were a breakthrough, but can no longer serve as a single trust anchor.
Quick context on where I’m coming from. I was the first to bring encrypted computation into blockchains: MPC with the Enigma whitepaper (2015), then TEEs with @SecretNetwork (2020), and today FHE with @fhenix. Each was right at the time; each has a place going forward. But only FHE can credibly be considered the end-game.
Why FHE? Because it’s the only cryptographic solution that’s simple to build on. Pure MPC is a Rube Goldberg machine — communication-bound and fundamentally limited. I wrote both my Master’s and PhD dissertation on MPC. If anyone has sunk costs in MPC, it’s me. And yet I’ll say it: it can’t take us further.
On the flipside, TEEs struggle under an expanding physical and micro-architectural attack surface. Still, projects employing TEEs like @SecretNetwork are the only real test cases we have for privacy at scale. Credit to them for building in public, patching, and reinforcing what will remain a key ingredient in privacy tech. It just can’t be the only (or main) line of defense. That must come from cryptography.
So are TEEs dead? No. That’s a lazy take. TEEs still deliver solid defense-in-depth and cheap privacy in high-trust environments. And for huge LLMs that need GPUs and low latency, GPU-TEEs may remain a pragmatic bridge (see @SecretNetwork AI Cloud).
With that said, all the anti-TEE privacy projects celebrating this “victory” should be more humble. How many of you are actually in production? That includes Fhenix, which I’m building right now. We all have a long way to go — and most of the posts I’ve seen from MPC/FHE/ZK projects are being disingenuous about their own privacy pitfalls.
Where we’re failing, as a community, is vocabulary. We need clear privacy levels — something as legible as @l2beat’s rollup stages, but for privacy: what’s protected, against which adversary, with which residual risks. I’m drafting a framework; if you want to pressure-test it or contribute, DM me.
The takeaway: treat TEEs as a component, not the foundation. Build toward FHE-first systems with MPC-hardened keying — and be explicit about the privacy level you’re actually delivering. That’s how we make programmable privacy for chains real.