Kimi K2 Instruct. Deepseek. Alibaba Qwen. Three of the world’s best AI model families. Financed and supported in part by the People’s Republic of China. World class coding and tool handling. I read the new paper from Anthropic about Chinese hackers supposedly using Claude Code to try breaking into different systems. The paper is very…
Summary In today's episode, I explain how to use AI to accurately tabulate data from unstructured PDF documents through a multi-stage process. Here's what this means for you. You can transform messy, complex contractual data into reliable tables without suffering from the common mathematical errors inherent in large language models. You'll also learn these concepts:…
Here’s how I use AI for health-related stuff. BIG HONKING DISCLAIMER: AI is not your doctor, and neither am I. The only person you should accept medical advice from is a qualified human healthcare provider who is knowledgeable about your specific situation. I am unqualified to give health advice. Health is something everyone has varying…
Summary In today's episode, I explain why ad value equivalence serves as a flawed measurement and how you can transition to more effective strategies. Here's what this means for you. You will learn to capture high-quality first-party data that proves your actual marketing impact through behavioral change. You'll also learn these concepts: why ad value…
Summary In today's episode, I explain why email click metrics often differ from Google Analytics page views. Here's what this means for you. You can stop chasing perfect data matches and instead focus on meaningful trends in your marketing performance. You'll also learn these concepts: why corporate security firewalls inflate click counts, how ad blockers…
OpenAI recently published a gigantic prompt library as examples for people to get started with AI. What’s the one thing in common with the vast majority of examples? They specify the need for context. Lots of context. The prompts themselves are incredibly generic, but clearly made to not overwhelm a beginner. Here’s an example: Write…
When it comes to reporting with data, especially quantitative data, in generative AI, you have to answer a critical question: Are we using the data for the report, or in the report? This seemingly pedantic question is vitally important. Will the data be used in the report itself, or will the data be used to…
Summary In today's episode, I walk through eight real-world data horror stories from marketers and map each one to the Six C's Data Quality Framework (clean, complete, comprehensive, calculable, chosen, and credible). Here's what this means for you. You walk away with a horror-story-tested approach to spotting data quality failures plus a workflow that audits…
Summary In today's episode, I examine the security implications and best practices for using AI note-takers in professional meetings. Here's what this means for you. You can protect sensitive information and maintain compliance by understanding the risks of cloud-based AI tools. You'll also learn these concepts: how regulatory and client requirements dictate tool usage, why…
The more specific a piece of data, the less useful it is in forecasting. This is a maxim of predictive analytics. Predicting audience behaviors based on the year they were born is useful. Predicting audience behaviors based on individuals’ exact birthdays is not useful. Which is why, in OpenAI’s newest paper on hallucination in generative…
