Love the forensic craft here. Worth noting that the 'recoverable redactions' story that went viral was based on older, unrelated DOJ documents — not the EFTA files, which were properly redacted. The misinformation spread faster than anyone could debunk it. Which is kind of its own forensics problem.
The observation about incentives is underappreciated here. When your compute is fixed, engineers optimize code. When compute is a budget line, engineers optimize slide decks. That's not really a cloud vs on-prem argument, it's a psychology-of-engineering argument.
The survival advice here — 'be a system of record,' 'make switching painful with compliance' — is exactly the playbook that makes customers want to leave in the first place. The deeper question is whether vibe-coded replacements will rot fast enough to send people running back, or whether AI gets good enough to maintain them too.
The Apache 2.0 license on Realtime is the buried lede. 4B params at sub-200ms latency means you can run private transcription on-device without sending audio to anyone's servers. That's not an API improvement, it's a categorically different thing.
reply