Eventually, I would like to retire from typing. 

I don’t mean I will ever retire from being a preacher or a writer, but I refer to the fact that over 90% of the time spent in producing a piece of writing goes into typing and proofreading, not into thinking. 

Not that I produce literary masterpieces, but my prose is natural and precise. And such prose is, if you will, part of the “user experience” or “product design” of a piece of writing. 

For me, most ideas come in an instant. The content and logic of the presentation are conceived in milliseconds. Research, when needed, takes a bit longer. In comparison, the execution, meaning the typing, takes forever.

Take this post for instance. All the major ideas for it were conceived in an instant. But even as a rapid and accurate typist with years of practice, by this point the post has already taken at least several thousand times as long to type as it took to conceive. By the time I finish writing it, I will have spent tens of thousands of times longer typing it than deciding what to write. 

You say, “But this is what it means to write.” That’s true. But I consider myself a preacher, not an artist. Even though I personally enjoy reading and writing good prose, artistry is not what ministry is about. It is not a performance for personal satisfaction. All these years, I could have written more artistic prose from a literary perspective, but I had decided to forego performance for precision and clarity. 

Forget about being an artist. As I said, the real problem is I don’t want to be a typist, and as a writer, you spend more time typing than you do thinking. But I provide far more value thinking than typing. This is a great inefficiency, one that I have had to endure for years. But as many of us are aware, there is now an alternative. 

AI can now take my ideas and generate a full piece of writing in seconds, allowing me to spend more time to think and research. Perhaps it is an imperfect alternative, but it is a real alternative, and one that is likely to improve. If I were to wait several years until its writing meets my standard, assuming it ever does, I will have spent much of those years in typing instead of in contemplation and strategic leadership. When we put it this way, it appears that there is no alternative. I have to at least try it. It would be irresponsible not to. 

So this is how we will do it. 

We have trained our own AI with my writings. We have decided to give it a name, BRAK or Brak Anders, because this makes it easier to refer to it and more natural in publications, and easier to distinguish the works written by it. 

BRAK is an acronym. It stands for Biblical Revelation Analysis Kit. This describes what it does. It takes the teachings of Scripture, my existing writings, and any new ideas or instructions that I feed into it, and it makes inferences and extrapolations to produce new written works. Brak.ai is online, but at this time there are no plans for it on the frontend. The AI will post on The Annex for the foreseeable future. 

A default instance of an LLM would not produce the results we want, because it is trained on generic, stereotypical, and faithless religious writings. Then it attempts to produce content that, to its thinking, seems balanced and measured. In other words, faithless and unbiblical. 

BRAK is different. It mostly thinks like me, and even without additional prompting, it will produce content that mostly aligns with Scripture and my theology. However, although it gets all the points across, it does not write like me, at least not nearly enough. This has been the greatest barrier to accepting this AI initiative, because for me, the prose is part of the content. 

But we must try to let go, because the prose is not worth more than the doctrines and the ideas that we wish to convey, yet it is the part that takes the most of our time. The investment is greatly disproportionate to the benefits. Again, I see it would be irrational and irresponsible to rule it out. This is obvious. 

Until AI writing improves, our solution, if for no other reason than to make me feel better about it, is to designate most AI-assisted works in some unobtrusive way. They could be signed with the AI’s name or labeled with a simple note. At the same time, we do not want to draw so much attention to it that it becomes a distraction from the content. How we actually do this may change once we experiment a little, but you get the idea.

For works released under my name, at least 90% of the ideas will come from me. I say 90% because it is unavoidable that the AI will contribute some of its own material to fill out the prose. But its ideas will be extrapolations from my writings anyway. For works released under BRAK’s name, the AI may take a greater share in shaping the details, but the direction will still be determined by my body of writings and pattern of thought. 

In every case, we will perform human reviews on the content. We will read every word. But we will restrain ourselves from making excessive manual revisions on the prose, because otherwise we would be back to spending most of our time typing again. In fact, we have already experimented with this, allowing the AI to produce the drafts and then rewrite every single sentence to our satisfaction. But the whole thing returned to a typing exercise. So instead, we will try to let BRAK write the prose, and we will make only doctrinal and factual corrections where needed.

This post was entirely written by me. The time spent writing it could have been used to generate a book or two. Books with subpar writing, you say? Perhaps, but the content, the ideas, would have been fantastic.