The massive growth of LLMs has been shocking to witness, especially considering my first experience with ChatGPT. A coworker gained beta access, and I used it to generate poems in the voices of various celebrities. It repeated words excessively and had a low word limit on its output. At the time, I saw it as nothing more than a cool toy.

I would have never predicted how powerful these models would become. There was certainly skepticism in online communities like Hacker News and Reddit, where many doubted the long-term viability of LLMs.

However, using these models for daily tasks and interacting with them at Google has left me with little doubt—they are here to stay. I find it highly unlikely that they will have no impact, or even a net positive effect, on the software engineering job market.

There is a lot of angst among software engineers, even within Google, about the trajectory of LLMs and how to plan for an uncertain future.

In an effort to foster a more productive conversation, I want to share my thoughts on these risks and potential strategies to mitigate or navigate them.

AGI Properties

When considering AGI and job automation, I find two key dimensions particularly interesting:

1. The time horizon—how soon AGI will emerge.
2. The strength of AGI—how capable it will be.

I define slow AGI as any system that takes more than five years to develop. This threshold feels appropriate because a five-year lead time, even for a major life change, allows for solid risk mitigation strategies to be put in place. If AGI arrives more quickly, protecting oneself and ensuring a comfortable future becomes significantly harder.

I define weak AGI as any system advanced enough to eliminate my ability to find work as a software engineer. While some might argue about whether such a system qualifies as true AGI, that distinction is irrelevant to me. What matters is its impact on my employment and life—not whether it meets a philosophical definition.

Strong AGI, on the other hand, is a system capable of fully automating all human jobs essential to sustaining human life.

Overall Approach

General Scenarios

Stay the Course: Software Engineer

Don't do something! Stand there!

Continue as a software engineer. Keep an eye on the ML space, but understand that you'll likely be employed long enough to carve out a decent living.

When the time comes, there will be a very large number of unemployed people. At that point, cultural shifts may make unemployment more palatable.

If cultural shifts don’t occur, there will at least be more information about the remaining playing field—what's left unconquered by AGI. You'll be able to plan better and find a more defensible position.

Find Alternative Employment

Never let a perfectly good crisis go to waste.

Choosing the right field to enter next involves many factors, but I’m drawing inspiration from the sources of monopoly power that businesses rely on to identify the most favorable characteristics.

Key Barriers to Entry:

• Legal Barriers – Does the government restrict entry into this industry, keeping wages high?
• Capital Requirements – How much money would a new competitor need to enter this field?
• Physical Location Barriers – Is this job "text to text" (text input for text output), or does it require a strong physical component, making it accessible only to those in a specific location?
  • Thomas Friedman, in The World is Flat, refers to these jobs as "anchored."

While I’m open to some additional education, I don’t want to spend an extensive amount of time retraining. Ideally, I’d reach a $100K salary within five years. However, I also want a field with a high earning ceiling, allowing me to continue retraining while employed to further increase my income potential.

I’ll use data from the U.S. Bureau of Labor Statistics to refine my search and explain my reasoning for favoring or rejecting specific paths.

Relevant list of occupations (see appendix for full data):

Unfavored

• Actuaries – Likely requires reeducation in statistics. While LLMs are currently weak at math, effective tool integration could eventually dominate this field.
• Computer and Information Research Scientists – Likely requires an MS in CS, a field I believe will become significantly less valuable.
• Information Security Analysts – CS-adjacent and potentially vulnerable to automation.
• Software Developers – My current field, which I expect to decline in value.
• Data Scientists – CS-adjacent and requires significant retraining in statistics.

Semi-Favored

• General and Operations Managers – Unlikely to land a C-level job anytime soon, making it a poor backup plan.
• Medical and Health Services Managers – The healthcare system has significant inefficiencies, particularly in administration and management (source). While this makes the field lucrative in the short term, long-term structural reforms could disrupt it.
• Management Analysts – These roles are typically filled by prestigious university graduates. While I would enjoy solving business problems, landing one of these jobs may not be realistic.
• Computer and Information Systems Managers – CS-adjacent but involves a more human-focused, deal-making component. I may be able to Slot into one of these roles Early on in a transition and then later transition away As I feel the field will continue to shrink
• Veterinarians – Requires extensive training and, in my opinion, is more challenging than nursing (with a higher risk of being bitten by an animal than a human). It meets all three key barriers to entry, but for personal reasons, I don't favor this route.

Highly Favored

• Financial Managers – A field I’m passionate about. It’s highly transactional but relatively insulated from overseas competition due to legal barriers. I’ve considered this my informal backup plan, though I may need additional finance training. This training could come in the form of further education, which would also serve as an additional barrier to entry. However, I believe LLMs could eventually handle financial analysis, making this a potentially short-term option.
• Registered Nurses – Pays slightly under $100K but offers growth potential into higher-paying roles like nurse practitioner or physician assistant. This profession aligns with all three barriers I’m optimizing for: it has licensing requirements (legal barriers), requires a significant financial investment in education (capital barriers), and is physically anchored.
• Nurse Practitioners – Requires a master’s degree but can be pursued as an advancement from an RN role, offering the same advantages.
• Physician Assistants – Requires a master’s degree but provides strong compensation. This role could be pursued as a follow-up to an RN and carries the same advantages.

Weak/Fast AGI: Financial Manager

I've already passed the Series 65 exam, and I believe I could establish myself in this industry relatively easily.

With a strong passion for economics and finance, this seems like the natural path if I get laid off and software engineering opportunities start to dwindle.

This field also offers the potential for a stable 9-to-5 job, which could give me the flexibility to study on the side if I decide to transition into nursing.

Slow/Strong AGI: Registered nurses

Returning to a two-year institution to earn an associate's degree in nursing is a viable option.

I can steadily increase my pay by continuing to take additional training. The stability and low likelihood of automation or an influx of new entrants make this field attractive. If the world moves toward fast but weak AGI, the resulting chaos will make a stable career even more valuable.

One significant tail risk is another pandemic, which could increase personal risk in this field. However, that remains a low-probability event in the grand scheme of things.

While there has been increasing discussion about robotics, I find it unlikely that automation will replace nurses in the near future. When people are in pain, human touch is essential—something robots won’t replicate effectively anytime soon. In the long term, I expect robotics to take over complex surgeries first, leaving lower-level nursing roles relatively untouched.

Even if robots advance to the point of automating nursing tasks, nurses form a large voting bloc and are legally protected through boards and licensing regulations. I find it unlikely that they would passively accept widespread automation without resistance.

Prepping: Doomsayer

Ahhhhh!

True strong AGI or ASI is on the horizon, leaving little time to worry about trivial matters like making money.

The very existence of the human species may be at stake, making it crucial to prepare for what were once considered tail-risk scenarios. Ensuring readiness for these potential disruptions is now a paramount priority.

While I’m not rushing to the nearest bunker with an emergency supply kit for the next few years, I will be taking some simple, common-sense measures to hedge against this potential outcome.

Appendix

Highest projected percentage growth

Most new jobs