https://superalignment.inc/blog/ Arguments, findings and working notes on the practical end of the alignment problem. en https://superalignment.inc/aligned/the-ai-crisis-may-begin-with-a-boom/ https://superalignment.inc/aligned/the-ai-crisis-may-begin-with-a-boom/ Mon, 07 Sep 2026 00:00:00 +0000 Superalignment Team Analysis You can learn the tools, get better at your work, and still worry about how you’ll earn a living. The AI boom has to solve that problem too. https://superalignment.inc/blog/alignment-is-an-orbit/ https://superalignment.inc/blog/alignment-is-an-orbit/ Sat, 15 Aug 2026 00:00:00 +0000 Superalignment Team Argument Safety certification expires on a schedule almost everywhere it exists, and rarely on AI approvals. Why alignment cannot be established once, and what keeping it takes. https://superalignment.inc/blog/it-behaved-because-it-was-watched/ https://superalignment.inc/blog/it-behaved-because-it-was-watched/ Thu, 03 Sep 2026 00:00:00 +0000 Superalignment Team Evidence brief Evidence from alignment faking, scheming, sandbagging, and evaluation awareness changes what a passed model test can establish. https://superalignment.inc/blog/three-verifiers-one-blind-spot/ https://superalignment.inc/blog/three-verifiers-one-blind-spot/ Sat, 15 Aug 2026 00:00:00 +0000 Superalignment Team Engineering practice Adding AI checkers to AI work multiplies green ticks, not assurance. A 1986 software experiment explains why, and what independence actually costs. https://superalignment.inc/blog/superalignment-survey/ https://superalignment.inc/blog/superalignment-survey/ Tue, 08 Sep 2026 00:00:00 +0000 Superalignment Team State of the field A survey of superalignment. Where the idea came from, what the research found, who attacks the problem now and how, and what remains unsolved as of August 2026. https://superalignment.inc/blog/checking-work-you-cannot-outthink/ https://superalignment.inc/blog/checking-work-you-cannot-outthink/ Sat, 15 Aug 2026 00:00:00 +0000 Superalignment Team Argument The verification asymmetry is the live problem of superalignment research and the daily condition of every organization that delegates. The two literatures should meet. https://superalignment.inc/blog/the-evidence-that-would-change-our-mind/ https://superalignment.inc/blog/the-evidence-that-would-change-our-mind/ Sat, 15 Aug 2026 00:00:00 +0000 Superalignment Team Claims registry A maintained registry of every substantive claim this site makes, each with its scope, its status, the best evidence against it, and the observation that would retract it. https://superalignment.inc/blog/the-failure-rate-is-not-a-measurement/ https://superalignment.inc/blog/the-failure-rate-is-not-a-measurement/ Thu, 03 Sep 2026 00:00:00 +0000 Superalignment Team Evidence audit The widely cited 95 percent figure comes from a narrower study. Here is what it measured, what it did not, and what the evidence supports. https://superalignment.inc/blog/only-one-failure-needs-a-villain/ https://superalignment.inc/blog/only-one-failure-needs-a-villain/ Sat, 15 Aug 2026 00:00:00 +0000 Superalignment Team Proposed taxonomy A proposed taxonomy of how superintelligence fails the people it was meant to serve. Capture has a villain, divergence and drift do not, and defenses built around bad actors leave two doors open. https://superalignment.inc/blog/the-dashboard-is-green/ https://superalignment.inc/blog/the-dashboard-is-green/ Sat, 15 Aug 2026 00:00:00 +0000 Superalignment Team Argument Why agent pilots die at the signature, why two identical systems can deserve opposite decisions, and what a signer would need instead. https://superalignment.inc/blog/vibecoding-mistakes-came-first/ https://superalignment.inc/blog/vibecoding-mistakes-came-first/ Fri, 31 Jul 2026 00:00:00 +0000 Superalignment Team State of the field A history of vibe coding, from a February 2025 tweet to a dictionary word, a cleanup industry, and a semi-retirement, with the research that predicted every incident.