7 posts
In previous posts, I've described LLMs as "helpful liars." This is useful mental model, but it doesn't really help answer the question, "how do you get them to lie less?" After using LLMs 15+ hours a day (plus a bunch of my own independent research on exactly this) I've developed a new mental model...
This post is the story behind the research. If you want the full paper with methodology, statistics, and raw data: Receipt-Gated Pipelines on GitHub. I caught three AI models fabricating security reports. Complete with CVE numbers. Severity ratings. Remediation advice. For vulnerabilities they...
This post is the story behind the research. If you want the full paper with methodology, statistics, and raw data: Structure Beats Scale on GitHub. This shouldn't have worked. I took a model that costs a tenth of a cent per call — Mercury 2, a diffusion-based reasoning model that nobody was talking...
The road to hell is paved with good intentions. People often intend to partake in physical activity. Yet, just as often, they find themselves frustrated at their failure to act on these intentions. Why is it that people remain inactive despite forming intentions to be active? A study by Conroy et...
One of the most interesting thing about a lot of habit-based interventions is that many don't do wildy better than other interventions in randomized control trials. They seem to do the same or maybe a tiny bet better. 12 weeks, however (the usual length of time for a nutrition or physical activity...
One area of debate in the behavior-research community is the interaction between habits and goals. Wood and Neal (2007) proposes an interesting model (above) that depends on three principles that have been fairly well established in prior research: 1) “Habits are cued by context,” 2) “Habit...
Ed Deci and Richard Ryan introduced SDT in 1985 as a framework for understanding motivation and how it can enrich our lives, formalizing it in 2000. Since then, many health interventions have been created with SDT at their core. In their 2012 paper Fortier, Duda, Guerin, and Teixeira, all of whom...