You have probably seen two headlines in the same week this year. The first: another round of tech layoffs, with AI cited as the driver. The second: software engineering openings are up sharply, and AI infrastructure roles cannot be filled fast enough. Both are true. The tech market in 2026 is not shrinking - it is splitting, and the split is happening faster than most people realize.

Q1 2026 saw roughly 78,000 tech jobs eliminated, with somewhere between 40% and 50% directly tied to AI automating the work. In the same quarter, software engineering listings rose about 30% year-over-year, and AI-adjacent roles grew even faster. If you are sitting on the wrong side of that split, you need a plan. If you are on the right side, you need to stay there.

This post is that plan. It is not motivational. It is a practical framework for evaluating your exposure, making the right moves, and doing it before your company's next reorg forces the question.

Part 1: Understand Which Roles Are Actually Exposed

Not every engineering role is at equal risk. The layoffs this year have concentrated in fairly predictable places, and pretending otherwise does not help you. Here is the honest read.

High exposure

Routine content and copy work. Marketing copy, basic technical writing, first-draft documentation. Tools are good enough that teams are doing this with one writer and AI instead of four writers.
Tier-1 customer support. The playbook is now: AI handles the first pass, humans handle escalations. This has cut entry-level support roles significantly.
Basic QA and manual testing. AI-generated test cases plus automated execution have reduced demand for manual QA.
Low-complexity frontend implementation. Building simple CRUD screens from designs is the work AI assistants do most competently. Teams that used to hire three junior frontend engineers are hiring one mid-level engineer with AI fluency.
Entry-level data analysis. Dashboards and basic SQL reporting are increasingly self-serve through AI query tools.

If your day-to-day is mostly in one of these buckets, you do not need to panic, but you do need a plan with a timeline measured in quarters, not years.

Moderate exposure

General backend engineering at companies without strong AI or infrastructure focus. Not going away, but compensation is flat and competition is fierce.
Mid-level full-stack roles at companies that are consolidating headcount. One engineer with AI tools doing the work of two is the pattern.
Project management and coordination roles without deep technical depth. AI tools are eating a lot of status-tracking work.

Low exposure (and often growing)

AI infrastructure and MLOps. Someone has to build the systems that serve models. Demand is enormous and supply is limited.
AI application engineering. Building RAG pipelines, agents, and AI-enabled features on top of foundation models.
Security engineering, especially application security and AI-specific security (prompt injection, model abuse, red teaming).
Platform and developer productivity. Internal tools, IaC, CI/CD, observability. Scales up as headcount per engineer goes up.
Senior system design and architecture. Judgment work that AI is nowhere close to replacing.
Specialized domains where domain knowledge is hard to automate - healthcare, fintech compliance, hardware, industrial control, safety-critical systems.

The pattern is not subtle: roles that are mostly "translate specification into code" are exposed. Roles that are "design systems, make tradeoffs, operate at scale, or apply deep domain knowledge" are not.

Part 2: Do an Honest Self-Audit

Before you decide what to do, get clear on where you actually stand. Three questions.

1. If you left tomorrow, how quickly could you land an equivalent role?

If the answer is "under a month" you are fine for now. If it is "I have not interviewed in four years and I'm not sure my skills translate," that is the signal to start moving.

2. In the last 12 months, have you shipped something that required judgment a model could not supply?

Things like: a system design that solved an ambiguous business problem, a performance investigation that required reasoning across multiple systems, a migration that required negotiating with stakeholders, a security review that caught something non-obvious. If yes, your work is already on the right side of the split. If no - if your last year has been implementing well-specified tickets - that is a red flag.

3. Can you describe, specifically, how your company's use of AI tools has changed your role in the last year?

If your answer is "not really," one of two things is happening. Either your company is not serious about AI (in which case your job security depends on them staying that way, which is a bet) or you are not engaging with the tools and you are about to be overtaken by peers who are. Neither is good. Both are fixable.

Part 3: The Four Moves That Actually Matter

There is a lot of career advice floating around right now. Most of it is noise. These four moves actually move the needle on your exposure, in rough priority order.

Move 1: Get real AI tool fluency, now

Not "I played with ChatGPT." Real fluency means you have integrated AI assistants into your daily workflow, you know where they are reliable and where they are not, and you can demonstrate productivity gains with concrete examples.

What to actually do this month:

Use Cursor, Claude Code, or Copilot for every coding task for 30 days straight. Track what works and what does not.
Learn to write specs and prompts that produce usable output the first time. This is a skill, and it takes practice.
Build one thing end-to-end using AI assistance and be prepared to talk about what you changed in the generated code and why.
Learn when to turn the AI off. That is part of the skill too.

Interviewers can tell the difference between engineers who use AI tools and engineers who parrot buzzwords about them. Actual fluency shows.

Move 2: Build production AI experience inside your current role

You do not need to switch jobs to move into AI work. You need to raise your hand for the AI-adjacent projects at your current company. Almost every engineering org has one, and the ones that do not are exactly the companies you should be leaving.

Specific things to volunteer for:

Integrating an LLM feature into an existing product (search, support, content generation).
Building or maintaining an internal AI tool for your team (code review bots, documentation generators, knowledge base assistants).
Setting up evaluation pipelines for AI features that are already shipped but not monitored.
Owning the infrastructure work to support an AI project - model serving, vector databases, cost monitoring.

Production experience beats certificates every time. "I shipped this AI feature" is the single most valuable line you can add to your resume this year.

Move 3: Level up your system design

The clearest pattern in the roles growing right now is that they require judgment. Senior and staff engineers who can design systems, reason about tradeoffs, and communicate technical decisions to stakeholders are not exposed to AI displacement. Engineers who implement tickets are.

If you have been coasting on implementation work, this is the thing to invest in. System design is learnable, and the delta between a mid-level engineer with strong system design chops and one without is enormous in this market.

How to build this skill:

Study the systems you already work on. Understand why they are designed the way they are, and what you would change.
Practice system design interviews even if you are not job searching. Articulating designs out loud is a different skill than doing them in your head.
Volunteer to lead the design review for your team's next significant change, and write the design document.
Read postmortems from public companies. Netflix, GitHub, Cloudflare, and Stripe all publish good ones.

Move 4: Build something visible outside of work

The downside of being "a good engineer at a company no one has heard of" in a market with 250 applicants per job posting is that you are indistinguishable from noise. A visible body of work - blog posts, open-source contributions, technical talks, a well-maintained GitHub - is a hedge against layoffs and a multiplier on your next job search.

This does not have to be a full-time second job. One decent blog post a month on something you actually know about is enough to compound over a year into real career insurance.

Part 4: What Not to Do

A few things that sound like good ideas but mostly are not.

Do not quit your job to take an AI bootcamp

Most of what the bootcamps teach, you can learn for free in 30 hours of focused work. The signal they send to hiring managers is mixed at best. If you already have an engineering role, your money is better spent on a GPU and time to actually build things.

Do not chase every new framework

LangChain, LangGraph, CrewAI, AutoGen, DSPy, LlamaIndex, Haystack, Semantic Kernel. You cannot learn all of them, and you do not need to. Pick one (LangGraph is a safe bet right now) and go deep. Depth in one framework plus strong fundamentals will serve you better than shallow familiarity with six.

Do not pivot to pure ML research

Unless you already have a strong ML background and access to top labs, pivoting into pure research is a multi-year project with brutal competition. Applied AI engineering - building things on top of existing models - has a much shorter path to employability and pays comparably at most companies.

Do not assume your current employer will retrain you

Some companies do retrain. Most do not. Plan as if yours does not, and be pleasantly surprised if it does.

Do not panic-apply to 400 jobs

Low-signal mass applications mostly produce rejection and exhaustion. 10 well-targeted applications, each with a tailored resume and a referral where possible, convert better than 400 cold submissions.

Part 5: Your Next 90 Days

If you do nothing else after reading this, do these things.

Week 1-2: Audit

Write down what percentage of your weekly work is well-specified implementation versus ambiguous problem-solving.
List every project you shipped in the last 12 months. For each, note what judgment you applied that AI could not have supplied.
Identify the single AI-adjacent project at your company you could plausibly volunteer for this quarter.

Week 3-6: Build fluency

Install Cursor or Claude Code, use it daily, keep a log of wins and frustrations.
Ship something small end-to-end using AI assistance. Could be a side project, an internal tool, anything.
Read one production AI system design write-up per week. Anthropic, OpenAI, Cloudflare, Vercel, Uber, and Netflix all publish good ones.

Week 7-10: Visibility

Write one blog post (or internal doc, if you prefer) about something technical you learned or shipped.
Update your LinkedIn and resume to reflect actual AI-adjacent work, not buzzwords.
Reach out to five former colleagues. Not to ask for jobs - just to stay in touch.

Week 11-13: Market-test

Apply to three roles, even if you are not seriously looking. The point is calibration.
Do a mock interview. Real or simulated, does not matter, as long as you get feedback.
Decide: am I on the right trajectory, or do I need to change companies this year?

Ninety days is not long, but it is enough to meaningfully change your trajectory. The engineers who will be complaining about the market in 2027 are the ones who spent 2026 waiting to see what happens. The ones who will be fine are the ones who treated this year as a planning year whether they felt threatened or not.

The Bottom Line

The 78K layoffs are real, and the 30% growth in SWE openings is also real. The difference between being on one side of that statistic or the other is not luck. It is mostly about whether you are doing the work that the market is paying more for, or the work that is getting automated.

Most engineers will not read this and do anything different. That is actually good news for the ones who do. The signal that separates engineers in this market is not credentials or prestige - it is intentional positioning, and a year of small moves compounds into a meaningfully different career position.

Pick one move from part 3. Do it this week. The rest will follow.

gitGood.dev helps engineers prepare for the interviews that matter in 2026 - AI mock interviews, 1,000+ practice questions, and coding challenges across all the roles actually hiring right now. The market rewards preparation. Start before you need to.