We need to talk about the AI hype cycle. Lately, it feels like every developer with a ChatGPT Plus subscription is calling themselves an AI expert, but the actual ML Engineer is currently navigating a much messier reality. Specifically, we are looking at a $200 billion investment bubble that assumes LLM technology will produce profits that, frankly, the math doesn’t always support. I’ve been around long enough to see the dot-com era’s ghost in these numbers.
In a recent conversation with Stephanie Kirmer, a Staff ML Engineer with nearly a decade of experience, she shared some blunt truths about how this role is actually changing. Consequently, the shift isn’t just about writing better prompts; it’s about the sociological impact of the code we ship and how we manage the “scut work” that used to eat our afternoons.
The Sociology of the ML Engineer
One of Kirmer’s most interesting takes is how her background in sociology shapes her engineering. In contrast to the “move fast and break things” mantra, she asks: “What are the social inequalities at play here?” and “How do different groups experience this tool differently?” This isn’t just academic fluff. In the WordPress ecosystem, I see this daily—how poorly implemented AI search features can alienate users who don’t fit the “standard” data model.
If you’re interested in how these data models actually apply to our niche, check out these 3 machine learning lessons for WordPress development. It’s a reality check for anyone thinking they can just “hook” an LLM into a site and call it a day.
Automating the Scut Work
Let’s be honest: writing unit tests and boilerplate code is a bottleneck. Kirmer notes that the day-to-day of an ML Engineer now involves using code assistants to handle this “scut work.” This frees up time for unique problem-solving, but it also creates a new type of technical debt in AI development if you aren’t careful.
Specifically, if you’re using an LLM to generate logic for a custom WordPress plugin, you need to ensure the output isn’t just “working” code, but secure, scalable code. Here is a simple example of how I might use a helper function to evaluate if an LLM-generated snippet meets our internal standards before it even hits a staging environment.
<?php
/**
* Naive evaluation of LLM-generated code blocks for specific banned patterns.
* Prefix: bbioon_
*/
function bbioon_validate_llm_snippet( $generated_code ) {
$banned_patterns = [
'eval(',
'base64_decode(',
'$_POST['
];
foreach ( $banned_patterns as $pattern ) {
if ( strpos( $generated_code, $pattern ) !== false ) {
return false; // Found a security bottleneck
}
}
return true; // Potentially safe to refactor
}
The $200 Billion Bubble
Kirmer argues that we are in a bubble, not because the tech is useless, but because the investment is disproportionate to the current utility. Therefore, companies are making outlandish promises to keep the VC wheel turning. As a senior dev, I’ve seen this before. When the bubble bursts, the engineers who survive are the ones who focused on solving actual problems rather than chasing hype.
Rebuilding trust with a skeptical public requires us to demystify the tech. We need to stop treating LLMs like magic and start treating them like the advanced statistical models they are. You can read the full interview with Stephanie Kirmer over at Towards Data Science for a deeper dive into these social themes.
Look, if this ML Engineer stuff is eating up your dev hours, let me handle it. I’ve been wrestling with WordPress and custom integrations since the 4.x days.
The Pragmatic Takeaway
The role of the ML Engineer is evolving away from pure data science and toward a mix of software engineering, sociology, and ethics. Don’t get distracted by the $200 billion noise. Instead, focus on building reproducible, trustworthy systems. Whether you’re working in Python or building custom blocks with the WordPress Developer Resources, the goal remains the same: solve the problem, don’t just ship the hype.
