Skyfall: How Tech Workers Survive and Thrive in an AI Era
When will AI replace us?
👋Hey friends, this is Leihua PhD from Tech Valley. I’m an avid content creator on AI, Technology, Data Science, Experimentation, Statistics, and Personal Growth. I appreciate your readership — let’s build a learning community together🍀🔥🤝
Also, please do me a favor and move this email to your primary folder.
Photo by Gerard Siderius on Unsplash
AI is taking over the world at a speed no one has seen before. AI follows a supercharged Moore Law and 2X its capacity every 6 months. As Charles Dickens in 2025 would put it,
This is the best of times, this is the worst of times for tech workers.
It’s the best because of the productivity boost and information accessibility. It’s the worst because of the fear of being replaced by AI.
Working in Silicon Valley, I’m surrounded by these two extreme emotions.
Will AI Replace Human Tech Workers?
I’ve been trying to replace my own work with various AI tools for the past 3+ years. It turns out, AI can handle 50% of the workload pretty well with minimal guidance. These tasks are well-defined, with clear context, constraints, and expected outcomes. Little to no ambiguity is expected.
One prime example is coding. After collecting all the product requirements and understanding the delivery, AI can write up an entire app with well-crafted prompt engineering.
However, AI is terrible at the other 50% of the job, which is ambiguous by nature and requires developers to wear multiple hats:
As a PM, to fully understand the business context and product value, which can’t and shouldn’t explicitly defined by a few prompts. Imagine prompting Google’s value as providing search services for users and expecting AI to fully appreciate Google’s product value — we know that would be a disaster.
As a domain specialist, to identify the commonality and uniqueness of the problem statement. Most of the time, we can refractor/reuse the existing code, but the precondition is knowing which parts we can refractor/reuse and which we can’t.
As a cross-domain generalist, data scientists think like engineers, and vice versa. This type of multi-disciplinary mindset is essential to high-caliber delivery.
As an effective communicator, to bridge gaps and bring everyone on board. We all know how difficult it is!
AI is not designed to answer these types of questions.
So, will AI replace human tech workers?
Partially. AI will replace some skills and job functions.
Now is the performance review season in tech companies. If I were ChatGPT’s manager, I’d probably rate it as a junior individual contributor (IC) and give it a SUCCESS rating. It functions like a junior engineer who requires a lot of hand-holding (via prompts) and needs help understanding business questions. It’s doing a decent job at what it’s designed for, but not enough to level up.
My prediction is AI will continue improving in the current causation ladder (ie association) but won’t move up to intervention and counterfactual any time soon. For folks who are new to causal inference, I’m referring to Prof. Judea Pearl’s Causal Hierarchy (1. association; 2. intervention; 3. counterfactual).
I don’t expect AI to develop human-level reasoning skills anytime soon. AGI remains a fantasy until its arrival. This is a great topic and deserves another post.
Why a SUCCESS Rating?
First off, Software Engineers and Data Scientists are more than just writing code. It’s an inaccurate oversimplification to categorize SDEs/DSs as just Programmers. A more accurate term is Developers.
Programmers rely on writing programs to solve problems.
Developers leverage all tools available to solve problems and bring business value.
“Developers” is a more accurate representation of our daily work. Coding is the last step in problem-solving, and there are miles to go before writing the first line of code.
Job Security
If AI replaces our job, it will start by replacing our skills. Some tech workers operate in established domains where questions and solutions are clearly defined. Job security in these areas doesn’t look so great because AI can learn and iterate fast.
Here are some established fields:
Building a Machine Learning model to predict customer demand
Running an A/B test and conducting hypothesis testing
Determining the default probability of a loan applicant based on their credit profile
…
We have accumulated so much prior information on these topics that there is little ambiguity left in the process. So, AI can leverage the rich body of knowledge and output results with high credibility. If you work in these fields, the job security might be at risk. Pivot early!
On the other hand, you are in a great position if you are dealing with ambiguous questions where the final solution isn’t perfect but works after trade-offs. On the non-technical side, you’ve identified key stakeholders and collaborated with them to get the job done. Sometimes you even wonder, "What the heck am I doing?" That’s a good sign. AI can’t take your job anytime soon.
At the end of the day, we need to make good decisions after weighing trade-offs, which are often unintuitive and elusive. The definition of "reasonable trade-offs" is a moving target. Let’s say there are three key factors in a decision. The simplistic way for AI to understand these factors might be assuming:
1 unit of Factor A = 1.5 units of Factor B = 1.8 units of Factor C.
However, the real world operates in a much more nuanced way, and trade-offs are rarely that clearly defined. And I’m glad they aren’t.
My Coping Strategy
You may have noticed there are fewer junior-level openings, and senior roles have become the new entry level. In an era of AI, everyone needs to be a 10X developer to stay relevant. AI has raised the bar for everyone.
Sharing a few personal coping strategies:
Faster coding != better coding. Deep dive into the non-coding aspects before jumping into code. Understand what problem I’m solving and how coding fits into the larger problem-solving process.
Adopt an interdisciplinary mindset. Wear multiple hats—engineer, data scientist, PM, communicator, and many more—and be comfortable operating across multiple domains. A multi-disciplinary mindset is a rare commodity and also your strongest advantage in the new era.
Expand your skill set. If a software engineer knows only one programming language and performs a limited set of tasks, AI is an immediate threat. Similarly, for data scientists—if your skills are limited to building ML models and running analyses in Jupyter Notebooks, AI can handle much of your work. Pick up new skills that AI lack.
Thanks for reading up this far.
When will AI replace human developers?
What are your coping strategies?
Please share your thoughts in the comments.
If you find this useful, please share this post and the Tech Valley newsletter with your friends and consider subscribing! Let’s build a learning community together.
I am happy to see you back in my feed! I missed your writing.
Regarding AI and the skills it has, we will remain competitive if we prioritize what I have heard called our Human Intelligence - the aspects AI cannot replace.