In episode 62 of The AI Fix, your hosts learn how AI models smash through CAPTCHA roadblocks like they're made of wet tissue paper - so much for humanity’s last line of defence. Meanwhile, we meet a bottle-flipping robot and call BS on a cartwheeling cyborg, Graham has a full-blown breakdown over traffic light puzzles, and Anthropic hires an AI psychiatry team to figure out why its bots are unhinged. Also: Microsoft helpfully tells us which jobs are toast, Vogue replaces models with AI-generated glam-bots, and a tiny homebrew neural net humiliates the big boys at Sudoku. Plus, Mark explores GEPA — the AI prompt optimiser that politely mutates your questions until they’re 35 times more effective than reinforcement learning. Episode links: AI model appears in Vogue . Humanoid robot can run, cartwheel and fist-fight . Nuclear waste cleaning humanoid robot nails 500 ml bottle throw test . 500ml Water Bottle Throw Test with ARMstrong Dex . Tweet by Jack Lindsey of Anthrophic . Anthrophic’s "AI psychiatry" team . Microsoft reveals the top ten jobs most (and least!) at risk from AI . This New AI is 100x Faster at Reasoning Than ChatGPT . Bias alert: LLMs suggest women seek lower salaries than men in job interviews . AI researcher ‘turns down $1bn pay offer from Mark Zuckerberg’ . Agent casually clicking the "I am not a robot" button . OpenAI’s ChatGPT Agent casually clicks through “I am not a robot” verification test . GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning . The AI Fix The AI Fix podcast is presented by Graham Cluley and Mark Stockley . Grab T-shirts, hoodies, mugs and other goodies in our online store . Learn more about the podcast at theaifix.show , and follow us on Bluesky at @theaifix.show . Never miss another episode by following us in your favourite podcast app . It's free! Like to give us some feedback or sponsor the podcast? Get in touch . Support the show and gain access to ad-free episodes by becoming a supporter: Join The AI Fix Plus! Privacy & Opt-Out: https://redcircle.com/privacy