I am less interested in reacting to individual releases than in understanding the conditions under which AI becomes genuinely useful and trustworthy.