Mark Pesce · University of Sydney · June 2026 Abstract When it becomes as hard to evaluate machine intelligence as it has always been to evaluate human intelligence, a threshold has been crossed. Human intelligence testing has been intractable for 150 years. Machine intelligence testing has become intractable in four. They are

Mark Pesce · University of Sydney · June 2026 tl;dr The failure of AI evaluations is itself the proof of the existence of AGI. AI evals are becoming intractable for the same reasons that measuring human general intelligence has been intractable for 150 years: the constructs resist decomposition, the benchmarks saturate,

Mark Pesce · University of Sydney · June 2026 Abstract AI evaluations are failing at the moment they matter most. As models approach general intelligence, evals must measure something we have never been able to measure in 150 years of trying. Simultaneously, benchmarks saturate through contamination and Goodhart effects, and the scope

Abstract The export controls on Fable 5 and the G7 governance proposals of June 2026 mark both halves of William Gibson's Turing Registry coming into being. Three capability tiers will emerge: unregulated, regulated, and restricted. But harnesses and model fusion routinely amplify models past their regulatory tier, making

Mark Pesce · University of Sydney · June 2026 Abstract "Superintelligence" conflates three distinct phenomena: projected (the ELIZA illusion), relative (genuinely superhuman within a domain), and actual (superhuman across all domains). Relative superintelligence is already here. Steve Yegge's "discernment horizon" names the operational consequence: past some

tl;dr Stack Overflow's new agent exchange lets agents both read and write to a shared code corpus, and that turns the Ralph loop into something collective: agents iterate on each other's work, hand it off, and improve it without anyone burning the tokens twice. The