Bruce Schneier and David Oppenheimer’s Opinion piece in the June 2025 edition of the CACM journal is a timely reminder that data integrity is required for AI systems to be trustworthy.
It’s a quick read and well worth your time:
Updated June 30th, 2025
A few days ago, I used GitHub Copilot to help create a custom dark theme for a MkDocs Material site.
Copilot did a good job interpreting CSS selectors, but the resulting theme’s accessibility was poor. Even when I nudged it toward better choices, some text ended up barely readable with low contrast to the background.
I've been writing more on LinkedIn lately and use Ulysses to keep a local record of those posts.
Recently, I decided to bring that content into my blog. At first, the idea of adding MkDocs headers, setting titles, organizing categories, cleaning up formatting, and splitting content into multiple files felt like a chore!
But GitHub Copilot, running in agent mode, turned the process into something fast, iterative, and semi-enjoyable.
Note: Copilot’s agent mode allows it to take actions on your behalf like creating project files, applying custom rules, and responding like an assistant to ask for clarification on steps to take. It does much more than code completion or even suggesting blocks of code.
What pleasantly surprised me was that I wasn’t dealing with code at all, and I plan to explore how this might support other documentation workflows, too.
The ability to describe what you want and get it without knowing the implementation details started with Fourth Generation Languages (4GLs) like SQL. SQL emerged in the 1970s and was commercialized alongside other 4GLs in the 1980s.
When someone claims we’re on the brink of AGI...
"Camelot! Camelot!"
... The reality?
"It’s only a model."
AI tools are powerful, sure. But true intelligence? That’s still a fantasy.
If you would rather not feed OpenAI or Google AI models after midnight …
Here is a good summary of how to use a robots.txt file to limit what OpenAI and Google will ingest from your website for training data.
Kudos to the authors for finding a valid use of the word gauche.
Read more here:
Microsoft has agreed to indemnify users of its Copilot service from copyright claims with its new "Copilot Copyright Commitment." This development may ease decision-makers' minds on potential copyright claims stemming from using the service.
What if an AI monitored whether you followed all rules and laws that applied to you?
This chilling and dystopian thought is the subject of a guest post by Jon Penney on Bruce Schneier’s blog (link at bottom).
Bruce Schneier has hit the nail on the head in his recent post on AI prompt injection attacks. Schneier feels it’s not possible to fully secure large language models (LLMs) against this kind of attack. Essentially, you can use AI to generate injection prompts, but read the article to learn more.
The ACM Technology Policy Council has published principles for generative AI. While good governance may not be as exciting as delivering new features, it is crucial for practitioners to act responsibly and thoughtfully in their work.