6 min read Mar 19, 2025

The current generation of consumer facing AI tools, known as Large Language Models (LLMs), continue to proliferate through society, used by folks from every stage of life, from office workers to school children. At a high level these tools are trained on a large collection of text-based content (such as every webpage on the internet and every book ever published; hard to get much larger than that) and sometimes further trained for a given industry (such as medical research, industrial design, law, etc.)

While it's true that LLMs are unable to reason or predict the future they are able to extract trends from the information they are trained on and they are able to output everything from bullet lists to essays to data that is easily parsed by computer programs.

Science fiction author Arthur C. Clarke once said that "Any sufficiently advanced technology is indistinguishable from magic." It's true that nascent technology is exciting, particularly when encountered for the first time. I recall my first ride in a self-driving car. Like all technology it proliferates society and adoption increases and before you know it light bulbs and radio are ubiquitously boring.

When LLMs first appeared schools went into full panic mode. Suddenly a middle schooler could type "write a three page book report about moby dick" into a prompt and paste the result into a Google Doc, print it out, and turn it in. Schools banned the technology and some sought tools to "detect AI writing," which turned out to be as accurate as flipping a coin.

Was such panic deserved? Possibly. The point of a book report is to test if a student read the required material and was able to comprehend what it was about and the use of an LLM tool defeated this test. That said, is it a beneficial assignment to ask a student to read a long book? I haven't a clue. Is it beneficial to ask a student to write an essay? More on that in a moment.

LLMs affect the business world as well. But let's take a step back. How are professionals expected to communicate with each other? Usually with empty platitudes to kick things off like "Hey! How's it going? How's the family? How are you doing?" Messages are expected to be written across several coherent English sentences, neatly organized into paragraphs, finally with some sort of signature. In the programming world we refer to this as boilerplate.

What is it that we are really trying to communicate? Usually a few short ideas that could be represented as bullet points, but that we need to fluff up with meaningless words so that we don't sound rude. Of course, this changes by culture and by language and is not applicable to many parts of the world, but it is definitely a thing in American English.

So now we see interactions with LLM tools that look a little like this:

HUMAN PROMPT:

Please write a professional and courteous email based on these topics:

• Joe's code quality is poor
• Joe needs to refactor his work instead of publishing Pull Requests so quickly
• Joe needs to stop annoying people to rush a review and merge his PRs
• I recommend Joe reads a copy of the book Code Complete

LLM RESPONSE:

Subject: Feedback on Code Quality and Development Practices

Hi Joe,

I appreciate the effort you’ve been putting into your work and your enthusiasm for getting changes merged quickly. I wanted to share some feedback that I believe will help improve the overall quality of your contributions.

I've noticed that some of your recent pull requests could benefit from additional refinement before submission. Rather than pushing PRs too quickly, taking more time to refactor and review your code internally can lead to higher-quality changes and smoother reviews. This will also help reduce the need for rushed feedback and rework later on.

Additionally, I’d like to encourage a more patient approach to the review process. While it’s great to be proactive, giving reviewers adequate time ensures a more thorough and thoughtful review, ultimately leading to better outcomes.

To further support your growth as a developer, I highly recommend reading Code Complete by Steve McConnell. It provides excellent insights into writing clean, maintainable, and well-structured code. I think you’ll find it valuable in refining your approach.

Let me know if you'd like to discuss any of this in more detail—I’d be happy to help.

Best,

{Your Name}

Of course, we're all busy, not only those who write email, but those who read it as well. So what does Joe do? He visits his friendly local LLM, of course.

HUMAN PROMPT:

Please convert the following long email from a colleague into a few bullet points:

Hi Joe,

…

Best,

Thomas

LLM RESPONSE:

• Appreciate your enthusiasm and effort in getting changes merged quickly.
• Suggest taking more time to refine and review code internally before submitting PRs.
• Encourage patience in the review process to allow for thorough feedback.
• Recommend reading Code Complete by Steve McConnell for insights on writing clean, maintainable code.
• Open to discussing further if needed.

Here we can think of the long email with meaningless fluffy padding as being the business speak protocol that office employees communicate with today. And we can think of the bullet points as how we actually think. And in this case Thomas the author of the email wrote a few bullet points from what was on his mind and had an LLM do the chore of converting it into business speak. Later Joe let an LLM do the chore of distilling business speak into short, actionable information.

The astute reader would have noticed that the translation was imperfect. Note the final bullet point that the recipient read where further discussion was offered. I can assure you that Thomas doesn't want to discuss it further.

As new technology shifts from new and exciting to boring and commonplace our ability to recognize this technology grows. The first few times I received physical junk mail letters in handwritten fonts and with pen marks I opened them thinking they were real letters. Eventually I realized they were mass produced mail created by machine and never opened another one again.

Do you remember those letters? The open rate was undoubtedly huge for the first generation. How exciting (to be a marketer!) Of course it tapered off quickly after that and I haven't seen one in several years. Me and everyone else can now spot those letters from a mile away. They've lost their magic and can no longer fool us. They've entered the uncanny valley; a place where things seem slightly amiss in an extremely off-putting way.

Text written by LLMs, with all the fluff and platitudes, will soon become so common and so easy to spot that it enters the uncanny valley for the majority of people.

I spot LLM-written content all the time. And I know you do too. Sometimes it's really annoying, like not until I'm already into the third paragraph of a webpage, but it happens. Sooner or later I realize I've eaten a bowl of fluff with no substance inside. And then I click back.

Anyway, I said that LLMs would change the world, but not in the way you'd think. How exactly will they change the world? LLMs are training us to hate fluff. Have you yet been accused that something you wrote was "generated by AI"? It'll happen eventually. We'll all hate fluff so much that, and this is the exciting part, we'll stop writing everything to be so fluffy.

Hell, kudos to you for reading this article. It's a six minute read. Most people can only manage one minute.

Remember those bullet points for sending emails? One day we'll just send bullet points as emails. We'll reach business speak protocol version 2.0. That which was verbose becomes terse. No more time wasted translating thoughts and injecting platitudes.

With much luck this will affect computer programming languages as well. Right now I can open a codebase, give an LLM that is trained on all of the world's open source code repositories a few bullet points explaining what I want a function to do, and it will translate those terse bullet points into verbose code. Why does it take a dozen lines of code to implement common programming patterns? It's because modern languages exist at an inconvenient level of abstraction. Perhaps it's time for a modern high level programming language to take off in popularity.

Once we get to the point where fluffy text is unnecessary, once we're all speaking business speak 2.0 and using the next terse generation of programming language with all the necessary patterns built in, we'll all be typing simple bullet points for everything, the world will be changed, and at that point LLMs may be entirely obsolete.