Here’s a hot take that might ruffle a few feathers: be careful dunking on TEEs just to make ZK/FHE look better. There’s a new attack from wiretap (dot) fail that demonstrates a practical DRAM-bus interposition attack that extracts SGX attestation keys and forges quotes. This is a real, concrete break we should take seriously. Does it mean we should punch down on TEEs? Probably not. A few loose thoughts: 1. Everything has vulnerabilities. Hardware, software, crypto libraries. None of it is magically immune. 2. TEEs have been around in various forms for much longer than ZK systems (TrustZone in mobile, DRM/payment chips, SGX since 2015). Because of this long history, researchers and adversaries alike better understand where to attack. That’s why we see a steady drumbeat of “another SGX exploit” headlines. 3. ZK systems, by comparison, are million-line libraries mostly with less than three years of production usage. They’re complex, evolving quickly, and their unknown unknowns are still waiting to be uncovered. 4. TEEs aren’t a static menu. SGX reflects one set of trade-offs. Other TEEs (your iPhone secure enclave, your bank card, your game console) make very different design choices, often prioritizing security over performance. The technology continues to evolve, and new, more secure options are already emerging. 5. Vulnerbilities will happen to ZK systems too. When they do, we shouldn’t cheer or gloat. Implementation is hard, and discovery of vulnerabilities is part of the maturation cycle. The bigger point: TEEs aren’t finished products. They’re a technology frontier, just like FHE or ZK. We don’t dismiss FHE because today’s benchmarks look bad. We evaluate it based on where it can go. The same mindset should apply to TEEs. They’ll probably never offer the same security profile as FHE, but they’ll likely remain far more performant and can serve as useful complements. And to be clear: this comes from someone who works across all three fronts. With EigenLayer and EigenCompute, we collaborate with TEE partners. With EigenDA, we work with ZK partners powering rollups. For slashing, we rely on reexecution consensus-based models where every Ethereum node replays the same code. Each of these approaches has different strengths and trade-offs, and each has a place. What shouldn’t become normalized is the idea that it’s fine to dunk on one camp as if it makes the others stronger. The verifiable pie is small. We should be working to expand it, not punching each other down. Build, disclose responsibly, patch quickly, and lift each other up.