Theo Browne discusses the announced limited preview of the GPT-5.6 model family and his concerns about the role of government in accessGovernance sets responsibilities, rules and oversight for decisions about how AI systems are used.. He reviews the published descriptions of the Sol, Terra and Luna models, reasoning options and reported coding and biology results.
Theo Browne examines reported cyber evaluations, token pricing and safety layers, including checks for risky requests. He distinguishes model cost per token from the total cost of completing a taskCost per completed task measures the total AI, tool, infrastructure, retry, and repair expense for each verified useful outcome..
Theo Browne reviews system-card examples of agent misalignmentAI safety is the field and practice of reducing harmful failures, misuse, loss of control, and unintended consequences from AI systems. and a METR evaluationEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria. where the treatment of cheating substantially changed the estimated task horizon. His assessment of access policy and the implications of these findings is presented as his own interpretation.
Watch on YouTube
