I learnt this week - 6th May 2025

The links today include an old obscure bug in Windows, browsing to think, technology non-predictions, AI apps, additional data, Geoguessr expert v AI, a simple question, a false law about headlines and a usable WinDBG debugger.

Why did Windows 7, for a few months, log on slower if you have a solid color background?

A nice explanation of how a simple missed condition (a solid background wallpaper) could cause an unexpected impact. Luckily this was fixed in 2009.

Browsing as thinking

The case for deliberately using vague or “bad search” to produce a broader set of results to browse for things of interest. I tend to be as specific as possible when searching but this could be an interesting exercise to try.

10 technologies that won’t exist in 5 years

A different approach to predicting the future. This is focussed on medical advances and makes predictions of what won’t be available in 5 years and which of them could potentially be achievable.

AI Horseless Carriages

Are we using AI in a way that is influenced too much by the way we expect software to work?

It seems like most AI “apps” are just revamps of existing software and involve wrapping AI prompts. The author here advocates that users should be able to edit those prompts to fine tune the way things work - and there are some really nice examples around handling emails.

YAGRI: You are gonna read it

I have recently been extending the data model of an application to include some of the proposed ideas in this article, especially around the dates.

o3 Beats a Master-Level Geoguessr Player

I’ve played Geoguessr a few times but definitely not to the level of the author of this article where he can identify a country by the type of roadside post.

This write up pits him against an AI model and it shows how he works things out and how the AI model does it’s reasoning as well. The results are pretty close.

A simple question that brought me alive

Always question why you are doing things - whether it is a meeting or an inconvenience. It is very easy to do things just because they have always been done that way or someone else thought it was a good idea.

Betteridge’s law of headlines

There is a theory that states that:

Any headline that ends in a question mark can be answered by the word no.

It is based on the assumption that if the answer was yes they would just have stated the fact. However a number of studies have dismissed this - nice idea though.

The Future of Crash Analysis: AI Meets WinDBG

I spent many hours using WinDBG for a previous role and it was painful. This approach using AI might finally make it usable!

That use of WinDBG actually got me a job. I interviewed for a role and was asked what I had been working on recently. I explained how I was using it to track down a particularly nasty issue. My new boss later told me that one reason I got the job was that if someone knows how to use WinDBG then must really know what they are doing. Hopefully I proved him correct.