Some thoughts on simple applications which turn out not to be so, game theory and plans for christmas, or early years resolutions.

Talking to your phone was supposed to feel futuristic… instead it feels like arguing with a drunk intern who speaks three languages badly. From mangling code terms to randomly answering in Portuguese when you spoke Spanish, mainstream voice-to-text is still a hot mess. Here’s the painful reality, and what I wish for the future.

An example illustrates how LLMs struggle with complex tasks unless broken down into simpler steps. By removing foreground letters, rotating the image, and re-adding the letters, Gemini demonstrates impressive image editing capabilities, making edits that are difficult to detect without watermarks. This highlights the potential of AI when tasks are properly structured.

AI agents are set to significantly enhance support experiences, providing efficient assistance compared to traditional human support. ElevenLabs offers an impressive example of conversational AI, while Vercel's AI Cloud allows for easy implementation of support solutions. The AI Gateway is a cost-effective option for experimenting with different models, streamlining the development process for support agents.

AI tools are evolving rapidly, but human brains will not keep pace. The feedback loop in human thought processes is crucial for creativity and understanding, as seen in activities like writing and driving. While AI can expedite tasks, it lacks the ability to replicate the nuanced thought processes that humans engage in, making the distillation of ideas a uniquely human capability for now.