Recent Posts

• AI Scheming Research Parallels Chimp Language Research (via)

  Published on 2025-07-18

  Tags: llms

  Interesting pre-print paper on the Arxiv comparing modern LLM “scheming” research to chimp language research in the 1970s. The authors point out the parallels between the two different areas of research and how AI “scheming” and safety research has many of the same characteristics that lead to the downfall of chimp language research. The main issues are that the evidence used is generally anecdotal, there aren’t control conditions, and the researchers play fast and loose with what they’re even testing without a rigorous definition of scheming.

  The paper states much more clearly some of the frustrations I’ve felt with AI doomers. They lack sufficient empirical validation for many of their core claims. It feels like AI doomers spend a lot of time theorizing about possible strategies for containing AI models or solving the alignment problem, without ever going out and just doing some solid research with at least something quantifiable.

  There are an abundance of “AI experts” out there like Eliezer Yudkowsky who haven’t done a lick of real work or research in any area of AI alignment or safety. At some point after 20 years you’d think you’d be able to put out at least a pre-print on the Arxiv with something approximating empirical evidence. But no, we keep going around and around with the same tired talking points about utility function misalignment. This paper is a call to do better. AI alignment might be an incredibly important issue with implications for the future of humanity. That should make the research all the more rigorous, which we just haven’t seen yet.

  I also wonder if there are parallels with the credibility revolution in economics? Economics used to suffer from a similar problem, lots of tidy and elegant theories without any research based on empirical evidence to back them up. Along came the credibility revolution (Wikipedia), which revolutionized economics into a much more rigorous and evidence-based field, with the result that economics has gotten broadly better as a whole. It seems like we need something similar for the field of AI scheming research. The field of computer science is already quite good about quantifying improvements, we need AI scheming research to follow that same pattern.

  Permalink

• What Caused the 'Baby Boom'?

  Published on 2025-07-15

  The common wisdom about the baby boom that was passed down to me was that once WW2 ended, all the lonely soldiers came home to their girlfriends, got married, and began a 15 year frenzied baby-making-fest. It’s one of those lullaby effect beliefs, it sounds reasonable until you say it out loud and think about the implications.

  Derek Thompson–previously of the Atlantic and now on Substack–has a better explanation. It mainly comes down to trends in technology making childbirth safer, labor saying machines that made having a family cheaper, and government policy making home mortgages easier to get.

  Permalink

• Don't Screw Me When I Leave, Because I Might Have Wanted to Come Back

  Published on 2025-07-13

  Often when you stop using a product or service, it seems like companies assume that you’re making a final decision for the rest of your life and treat you accordingly. Off the top of my head the ones I’ve recently run into, Adobe charges you the rest of your subscription fee as a cancellation fee and Square Space will make you wait a minimum of 24 hours if you want to move your domain off their site.

  The assumption that I’ll never use a company’s product or service again seems incredibly faulty. Take Adobe Photoshop for example. I got a short subscription to try it out for editing some photos I had. It seemed like a decent enough piece of software and I didn’t have any major complaints. However, when I finished using it after a few weeks and went to cancel my subscription I was met with a hefty cancellation fee. That just leaves a terrible taste in my mouth. Canceling my subscription in no way means that I’ll never need to edit another photo or that if I did want to edit more photos in the future that I wouldn’t have used Adobe Photoshop again. However, screwing me over on the way out makes certain that I won’t ever use your product again.

  Even if we were to assume that I really would never need to edit another photo in my life, that still ignores word-of-mouth advertising. Because not only will I never use Adobe Photoshop again, I have no interest in recommending Adobe Photoshop to any friends or family who need some editing software. This kind of behavior, milking the customer for every last penny you can squeeze from them on the way out, seems like it may be beneficial in the short term, but quite harmful in the long run. This may simply be a symptom of a bad corporate structure in these companies where execs are prioritizing quarterly results over a good name and long term growth.

• METR Impact of AI on Developer Productivity (via)

  Published on 2025-07-11

  Tags: llms

  Simon Willison’s comment on Hacker News sums up my feelings well:

  “Here’s the full paper, which has a lot of details missing from the summary linked above: https://metr.org/Early_2025_AI_Experienced_OS_Devs_Study.pdf

  My personal theory is that getting a significant productivity boost from LLM assistance and AI tools has a much steeper learning curve than most people expect.

  This study had 16 participants, with a mix of previous exposure to AI tools - 56% of them had never used Cursor before, and the study was mainly about Cursor.

  They then had those 16 participants work on issues (about 15 each), where each issue was randomly assigned a “you can use AI” v.s. “you can’t use AI” rule.

  So each developer worked on a mix of AI-tasks and no-AI-tasks during the study.

  A quarter of the participants saw increased performance, 3/4 saw reduced performance.

  One of the top performers for AI was also someone with the most previous Cursor experience. The paper acknowledges that here:

  However, we see positive speedup for the one developer who has more than 50 hours of Cursor experience, so it’s plausible that there is a high skill ceiling for using Cursor, such that developers with significant experience see positive speedup.

  My intuition here is that this study mainly demonstrated that the learning curve on AI-assisted development is high enough that asking developers to bake it into their existing workflows reduces their performance while they climb that learing curve.”

  LLM driven tools are incredibly powerful, anyone denying that is huffing copium, but the field hasn’t stabilized enough for devs to really grok the tools and how they should fit into their workflow. There’s a new model being released every month, each with a slightly different prompting style and behavior that subtly impacts their performance. Before you have a chance to really understand what type of prompting and context leads to the best performance for a model, a new one has already been released and you have to start over.

  Like any tool, it takes time before the speed increases kick in. I can say from personal experience that initially when I switched to Vim my ability to type and edit text took a significant hit. I probably didn’t exceed my previous speed for weeks. The same is true of using git. I initially spent hours and hours reading the Git book, and trying to perform simple commands in git before using the tool was actually a productivity boost rather than a hindrance when trying to write code.

  Saying: this tool takes time to master and use effectively is in no way a cop out. It would be surprising if that wasn’t the case. I would be willing to bet that the majority of power user tools have a steep initial learning curve.

  Permalink

• Gowall Wallpaper Utility

  Published on 2025-07-08

  Neat little tool I found to convert wallpapers to a colorscheme. It’s bugged me for a while that so many of the beautiful wallpapers I like don’t match the colorscheme of my apps and Sketchybar status bar. I already used it to convert all my wallpapers to my preferred colorscheme Tokyo Night. It helps to improve the cohesiveness of the colors.

  Permalink