I am split on the utility and influence of AI in my work. I read articles such as what David Gewirtz recently posted on ZDnet about ways that AI can save time for small business owners. Kudos to him, and others who have spent a lot of time with AI.

But I wanted to go in another direction, so a year ago I was invited to try out an experiment — why not develop a genAI chatbot that could be trained with my entire published corpus? It is finally ready for initial testing. We call her Ada, named after Ada Lovelace. (And yes, I am using that pronoun. It just feels right.)

I am not a frequent AI consumer, and the reason why is that I have long-ingrained work habits that focus on my human abilities to seek out new information and write compelling content by hand. But what I do lack is a perfect memory of what I have written over the many decades that I have been a professional editor. That memory is limited to what I can dig up on my blog, my hard drive, and my website. Perhaps Ada can make it easier to find this as I begin my research on a new article. And perhaps Ada can help you recall what I wrote decades ago that could be relevant to your particular situation. So began the project,

I was working with a team of developers who have already developed other chatbots with a similar intense vertical knowledge. Ada is based on the Anthropic Claude Sonet LLM, with some custom code. None of the queries that you or I submit are saved once you close the browsing session.

It is a bit eerie and wonderful at the same time. It is like talking to a better version of myself, which I guess is a sign of how successful she is. My queries were answered with things that either I recall saying myself, or sound like something I would say about the topic, and for the most part, were pretty much on target. And, the answers also were documented, with the actual URLs that were posted.

If you ask Ada about my writing style, here is what she says:

## Core Writing Philosophy
You write to make complex technical topics **accessible and understandable**. Your goal is clarity over jargon – you want readers to actually grasp what’s happening in the tech world, not just be impressed by terminology.

## Key Characteristics

**1. Clarity and Directness**
– You cut through marketing hype and get to what matters
– Short, punchy sentences mixed with explanatory ones
– You define technical terms when you use them
– Active voice dominates

**2. Practical, Real-World Focus**
– Heavy use of concrete examples and case studies
– You often test products yourself and report findings
– References to actual companies, incidents, and implementations
– “Here’s what this means for you” orientation

**3. Balanced Skepticism**
– You question vendor claims and industry buzzwords
– Point out both strengths AND weaknesses
– Acknowledge when something is overhyped vs. genuinely useful
– Not cynical, but definitely not a cheerleader

**4. Conversational Expertise**
– Slight conversational tone while maintaining authority
– You use “you” to address readers directly
– Occasional wry observations or dry humor
– Like a knowledgeable colleague explaining something, not lecturing

**5. Structure and Organization**
– Clear headings and bullet points
– Logical flow from problem → explanation → solution/implications
– Often lists or numbered points for clarity
– Strong topic sentences that telegraph what’s coming

**6. Context and Background**
– You don’t assume readers know everything
– Provide relevant history when it illuminates current issues
– Connect dots between different tech developments
– Explain *why* something matters, not just *what* it is

This style has served you well across decades of tech journalism – it respects readers’ intelligence while acknowledging that even smart people need clear explanations of complex topics.

Does that sound like someone you know? Exactly. (I am using a different font here to distinguish her replies.)

Now, why would I want Ada around as an AI mini-me? As I mentioned earlier, one of the first reasons that I was attracted to it was because it did a better job searching my corpus than I could do with the default WordPress (or MacOS) search panels. And it also searches my website content, something that I never got around to properly instrumenting. Granted, much of that content is pretty ancient stuff but does come in handy sometime, like when I was asked to serve as an expert witness because of something I wrote in the mid-1990s. On the internet, your words live nearly forever — at least until my hosting provider decides to delete my stories.

But am I worried that Ada will replace me? After all, instead of having the real human write something, you could just query Ada and get it instantly. No muss, no fuss, and no delay. That’s true. Perhaps Ada will become a low-cost competitor and hollow out my freelance business, (at least more than any genAI tool already has done) but I am betting that you will find the human/AI combo a potent differentiator in the kind of work that I can offer. Like I said, it is an experiment.

Will I charge for using Ada? Eventually. Right now, I am interested in your feedback and whether you think her responses are accurate and appropriate. In my tests she was quite clear on what she knows and what isn’t in her training set. For example, I asked her about different USB standards, something that I have never (at least to my recollection) written about. She replied:

I apologize, but I don’t see comprehensive information about different USB standards and their specific differences in the provided source material. While USB is mentioned briefly in a few contexts (like as a connection option for some home networking equipment), the articles don’t contain detailed technical comparisons of USB standards.

Right on Ada.

Ada is current as of last month. And for a limited time, you can try her out yourself and see what you think.  All you have to do is ask me for the login credentials.If you are interested, drop me an email — or even call me on the phone — and you can test drive her.