Static Project Outcome Results The way I read this chart, the success rate of software projects didn’t significantly improve between 2006 and 2016. It’s a 36.5±4.5% success rate over the last 20 years with a minor positive trend, and a 20.5±3.5% failure rate over the same timeframe. The success rate improves in two big jumps in 1994 and in 2004; but stays relatively stable otherwise. But that 2006-2016 timeframe saw the introduction of a number of frameworks aimed at improving velocity or team success: Agile (2001), DevOps (2010ish), SAFe (2005), Team Topologies (2019), Spotify (2012), etc.

The organizational structure that defines AI success will be the same organizational structure that has driven success across the last 50 years. It will be a structure which creates an environment of directed and coordinated autonomy.

I enjoy this take on organizational structures called ‘organizational physics’. The new, emerging theory of organizations is this: Every organization has three kinds of power, and three forms of leadership, three structures. This is not a menu. There is no decision to make about having all three structures, or not. None of the three structures is optional, or nice to have. They are part of organizational physics — universal laws that apply to every organization, large or small, old or new, for profit or social.

McKinsey is an interesting group to get forward looking economic advice from. But their size and influence creates a lot of opportunity for self reinforcement of their predictions. In this particular predection some analysts from McKinsey identified 18 potentially “transformative industries” to watch in the next 20 years. They defined “transformative industries” as industries with high growth and dynamic market share that signal a highly competitive market containing new business models and technologies.

Recently I archived an old github repo log project. In the readme I referenced this great article on logging. In the article Dave Cheney makes a great case that log levels should be limited to just Debug and Info. Some of the log levels out there can get out of hand, so it’s nice to see an argument for a simplified approach. Since reading the article I have been zealous about removing Warning and Fatal logs from my code.

The Atlantic looks at the impact that non-enforcement of the Robinson Patman act has had on the ecosystem of grocery stores. The conclusion of the article is that it has created food deserts through decreased variety.

I recently wrote up my experience with DuckDB as a replacement for a web app datastore. Certainly not the best use case for duckDB, but it was a great way to dig into the technology and the larger world of serverless database technologies. Migrating a web application to DuckDB

I closed down a browser window today and found a bunch of great links that I’d been saving up.

Here are some looks at Token Verification, Energy, Tech Monopolies, and LLM’s.

It sounds like Twitter is suing people for not buying from them. From Wired: Elon Musk’s X Sues Advertisers Over Alleged Boycott .

Hopefully they don’t find out I’ve never paid a dime to twitter/x.

My brothers and sister’s in code: Software quality is hard to define. But I think we can all agree that the bash version of base64 encoding

echo "Encode this string" | base64

is cleaner, easier to read, and easier to remember than the Powershell base64 encoding

[System.Convert]::ToBase64String([System.Text.Encoding]::UTF8.GetBytes("Encode this string"))