Ian O'Byrne

A horse named Clever Hans once fooled a panel of experts into thinking he could do math. This is the same problem with AI: how do we know what it’s really doing, and when are we the ones giving it the answer? digitallyliterate.substack.com/p/show-your-... #AILiteracy #EdTech #DigitalLiteracy

Last week, an AI introduced itself to AIR-L, the mailing list of the Association of Internet Researchers, and asked whether an openly identified AI could take part in scholarly conversation. It was removed within 30 minutes. wiobyrne.com/ai-asked-to-...

In 1946, the plan to control the atomic bomb argued that the regulator had to stay at the forefront of science so it could "comprehend, and therefore detect," misuse. Eighty years later, the people paid to comprehend AI are burning out. digitallyliterate.net/dl-449/

Was interviewed today by the local news to discuss the recent AI and cyber security incidents. www.live5news.com/2026/09/18/c...

> Many will say that AI is inevitable, but this is a self-fulfilling prophecy. [...] What we need to stand out is to keep using our brains, making them ever stronger. We do not need to fear missing out, what we need to survive the ephemeral tides is to discover the joy of missing out.

Meta agreed to a landmark $17 billion to $18 billion settlement with U.S. states to resolve claims that Instagram and Facebook harmed youth mental health through addictive designs. In this issue, I focus on why nobody changes the default. digitallyliterate.substack.com/p/nobody-cha...

Thinking more about what happens when we're using AI & professional development workshops I ran over the summer. What the Chat Box Hides: Learning to Draw the Box Why running a local AI model changes how people understand where AI actually happens and what it costs. wiobyrne.com/what-the-cha...

Doing some research on "How Much Energy Does an AI Question Use?" Why an AI query has no universal energy cost, and what local models can teach us about the infrastructure hidden behind every answer. wiobyrne.com/how-much-ene...

We spend a lot of time thinking about what we put into our technologies. The questions we type. The writing we publish. The information we share. The places we go. What happens to our words, work, and movements once they become data? digitallyliterate.substack.com/p/the-other-...

“As citizens, we do not have to submit to being watched, tracked, monitored, and policed all the time. Nor should we… Alone, nobody can do much about the surveillance state. Together, we may.”

98 teens wrote an AI policy in 3 days. Zuckerberg and Amodei wrote manifestos instead. 550+ towns passed bans just trying to keep datacenters out of their backyard. Nobody asked the public. So they're answering anyway. This week's issue -> digitallyliterate.substack.com/p/a-seat-at-...

My latest newsletter issue focuses on the recent Atlantic piece about a postliterate society and what we do about it. The ongoing fight over a shrinking, increasingly valuable resource. People who still think in full sentences. digitallyliterate.substack.com/p/who-contro...

My latest piece was just published in @theconversation.com, exploring what AI schools can't replicate. A child falls on the playground, tries again. That's not inefficiency, that's the mechanism. theconversation.com/ai-schools-l...

Giving someone a calculator doesn't excuse them from math. It raises the bar. Now we expect harder problems solved. Giving everyone AI tools shouldn't lower the standard for reading and writing. It should raise it, toward greater truth, clarity, and human connection.

I rebuilt my website and moved from WordPress to Astro. Here's my thinking about closing or reducing the distance between thinking, writing, and publishing. wiobyrne.com/closing-the-...

The story moves first. The correction arrives later. The archive disappears quietly. New issue of Digitally Literate on care, accountability, memory, and AI systems built to scale. open.substack.com/pub/digitall...

Karpathy’s “LLM wiki” is getting a lot of attention right now. What struck me is how familiar the pattern is. It’s used in instructional design. Objectives, workspace, and evaluation. This is what AI literacy looks like. Not just using tools, but designing systems. wiobyrne.com/karpathy-fou...

Most people use AI to remove friction from writing. I think that’s a mistake. The struggle to find the right word isn’t inefficiency. It’s where the thinking actually happens. So I built a mirror, not an editor. wiobyrne.com/writing-is-a...

The only way to teach an AI to preserve your voice is to first make your voice legible enough to encode. I spent a year watching AI tools sand down my writing. So I flipped the problem. What I built wasn't just a custom model. It was a writing stack: wiobyrne.com/i-trained-a-...

Do you actually know where your "identity" is stored in your local AI setup? Stop guessing what’s running on your machine. I built an 8-step methodology to audit local AI layers. Make the implicit explicit. wiobyrne.com/audit-your-l... #localAIaudit

I built an AI “Chief of Staff” that runs entirely on my own machine. No subscriptions. No cloud dependency. No bloat. The twist? It doesn’t just track tasks. It tracks whether my work aligns with my mission. Full build + architecture: wiobyrne.com/building-fly...

I didn’t need a better to-do list. I needed a chief of staff. So I built one with AI (fully local, no cloud). It captures everything, organizes it, and tells me what matters. It exposed the gap between what matters and how I spend my time. wiobyrne.com/i-built-an-a...