This newsletter you couldn’t wait to open? It runs on beehiiv — the absolute best platform for email newsletters.

Our editor makes your content look like Picasso in the inbox. Your website? Beautiful and ready to capture subscribers on day one.

And when it’s time to monetize, you don’t need to duct-tape a dozen tools together. Paid subscriptions, referrals, and a (super easy-to-use) global ad network — it’s all built in.

beehiiv isn’t just the best choice. It’s the only choice that makes sense.

Start free today. No credit card required

Hello, fellow humans!

Last time, we walked through what the AI market correction most likely looks like—a series of acqui-hires and asset liquidations to transform AI from a product category to an infrastructure layer. We talked about the three tiers of tech workers and who's positioned where. If you missed it, the key insight was this: the stratification isn't about technical depth, it's about transferability and sustainability of value creation.

Today we're getting tactical. Where are the actual opportunities? How should you position yourself based on where you are in your career? And here's the contrarian take I promised: why might a market correction actually accelerate real-world AI adoption and create better career opportunities for more people?

Let's start with the central thesis that changes everything about how you should be thinking about your positioning.

Here's what happens when building LLMs stops being the goal: companies shift their focus to solving actual problems with existing models. And that's when you get boring, profitable AI infrastructure that transforms industries without making headlines.

Think about the cloud transition. AWS didn't win because they had the most advanced data center technology. They won because they made it boringly reliable and easy to use. The same dynamic is about to play out with AI. The winners won't be the companies with the most impressive model benchmarks. They'll be the ones that make AI capabilities accessible, reliable, and economically viable for real business problems.

This shift creates four distinct opportunity zones, and understanding which one matches your skills and interests is critical for positioning yourself effectively.

Every enterprise has workflows that could be 40 percent more efficient with AI, but they don't need custom models. They need someone who understands their business and can implement AI solutions using commodity models. This is plumbing work, not research, and that's precisely why it's valuable and sustainable.

Let me give you a concrete example. A mid-size insurance company needs to process claim documents faster. They don't need someone to train a custom model on insurance claims. They need someone who understands insurance workflows, knows enough about LLM capabilities to identify what can be automated, can architect a system using existing APIs that's reliable and maintainable, and can measure the actual business impact in terms of processing time and error rates.

The person who wins this engagement isn't the one with the deepest knowledge of transformer architectures. It's the product manager who can scope the project realistically, understand the change management required to get adjusters to trust the AI system, and deliver something that actually gets used rather than gathering dust. Or it's the senior engineer who can build reliable systems with proper error handling, monitoring, and fallbacks rather than impressive demos that break in production.

The compensation trajectory here is $200,000 to $350,000 depending on domain and seniority. That's not peak-hype money, but it's sustainable because you're solving real problems that generate measurable ROI. More importantly, these roles exist at thousands of companies, not just a handful of well-funded AI startups.

The positioning strategy if you want to move into this zone: start picking up projects at your current company where you wire AI capabilities into existing workflows. Build a portfolio of boring wins—20 percent efficiency gains repeated 10 times is more impressive than one moonshot that never shipped. Learn to speak the language of business outcomes, not technical capabilities.

As foundation models commoditize, value moves to last-mile applications in specific industries. Generic AI tools are 60 percent solutions. Vertical applications can charge premium prices by solving the last 40 percent that requires domain knowledge.

The key insight here is that domain expertise is harder to acquire than technical skills. A radiologist who learns enough ML to be dangerous beats an ML engineer who tries to learn radiology. The medical knowledge takes 10 years to develop. The ML skills can be picked up in 18 months if you're motivated and have the right background.

This is where I think some of the most interesting opportunities are for people in that middle tier we talked about last time. If you've spent 5 to 10 years in legal tech, medical imaging, agricultural planning, supply chain optimization, or financial services, and you've spent the last 2 years learning AI, you have a combination that's genuinely rare and valuable.

The tactical move here is to target early-stage vertical AI companies that are actually solving real problems in your domain. These companies will have unit economics that work because they're charging for outcomes, not API calls. Look for companies where the founders have domain expertise themselves—that's a signal they understand the 40 percent problem, not just the 60 percent that generic models solve.

Compensation here tends to be equity-heavy, but unlike pure AI infrastructure companies, these businesses have a path to profitability. You're trading some cash compensation for meaningful equity in a company that might actually be worth something in five years.

This is the opportunity that almost nobody is talking about, but it's going to be huge. The AI productivity gains that everyone promised only happen if workers actually adopt the tools and change their processes. Most companies are terrible at this organizational change work.

Think about what happened with previous technology transitions. Companies bought enterprise software and then wondered why adoption was low and ROI was disappointing. The problem wasn't the software—it was the lack of training, change management, and workflow redesign. AI is going to be exactly the same, except the stakes are higher because AI capabilities are more powerful and the potential for misuse is greater.

Companies are going to need people who can upskill existing workforces to use AI tools effectively. This is training, but it's also change management, workflow analysis, and process redesign. It requires a combination of technical credibility and communication skills that's surprisingly rare.

The people who win here are former engineering managers who know enough about the technology to be credible but have developed strong communication and teaching skills. Technical product managers who've run cross-functional projects. Developer advocates who've built communities and created educational content. If you've ever been the person who explains complex technical concepts to non-technical stakeholders and actually gets through to them, this is your zone.

Compensation tends to be $150,000 to $250,000, which is less than pure engineering roles, but the work is more stable and often more satisfying. You're helping people become more effective rather than chasing the next model benchmark. There's also a clear path to building a consulting practice if you want more autonomy and upside.

Here's the dynamic that's going to play out: as proprietary models commoditize, open source becomes the default for 80 percent of use cases. Every company will run AI workloads. That means open source model serving, fine-tuning tooling, and evaluation frameworks become mission-critical infrastructure. Mission-critical infrastructure needs maintenance.

If you're contributing meaningfully to projects like Hugging Face, vLLM, MLflow, or the ecosystem around open models, you're building a reputation that will translate into multiple career paths. You might get hired by cloud providers who need that expertise in-house. You might build a consulting practice helping enterprises deploy and maintain open source AI infrastructure. You might get acqui-hired by a large enterprise that decides they need your specific knowledge to run their AI systems.

The compensation here is variable—it depends on which path you take—but the key is that you're building leverage through public work. You're not dependent on any single company's success. You're positioning yourself as infrastructure for an ecosystem.

The tactical move is to pick a project that's genuinely useful and that you can contribute to consistently over time. Don't try to spread yourself across 10 projects. Pick one or two where you can become a recognized expert. Write documentation. Help with issues. Contribute meaningful code. Build relationships with other maintainers.

Let me give you concrete actions based on where you are in your career right now.

If you're currently at an LLM company, you need to read the warning signs. If your company just raised a huge round but unit economics haven't improved, you have 12 to 18 months. Don't wait until layoffs are announced. Start documenting your wins in language that translates outside the AI hype cycle—talk about cost reduction, efficiency gains, revenue impact, not model perplexity scores. Build relationships with potential acquirers by attending their conferences and contributing to their open source projects. Most importantly, develop domain expertise in parallel. Pick an industry and learn it deeply enough that you can tell the story of how you're an industry expert who happens to know AI rather than an AI person looking for a domain.

If you're at a tech company that's adding AI features, you're actually in the best position. You have stable employment while building AI skills in the context of real business problems. Volunteer for AI integration projects but focus obsessively on the business outcome, not the technology. Learn the economics—what does it cost to run inference? How does that impact margins? What's the threshold where AI features become profitable versus just impressive? Build a portfolio of wins that demonstrate you understand both the technology and the business.

If you're trying to break into AI from another field, don't try to compete on pure ML credentials. You will lose that battle to people with PhDs and years of experience. Instead, take your 10 years of domain expertise in finance or healthcare or logistics or whatever you know deeply, and learn enough AI to be dangerous. Build demos showing AI applied to real problems in your domain. Target vertical AI companies or enterprise adoption roles where your domain knowledge is actually more valuable than deep ML expertise.

If you're early in your career, the commoditization is actually good news for you. You don't need to understand transformer architectures from first principles to build valuable AI applications. Focus on building products using existing AI APIs. Develop taste for what AI is actually good at versus what's just hype. Pick a domain and go deep—become the AI person in a specific industry rather than trying to be a generalist AI expert.

I’m seeing more people say that a funding correction might actually be the best thing that happens to AI as a field, and I have to agree.

When building LLMs stops being the goal, resources flow to solving actual problems. When companies can't raise infinite capital on the promise of AGI, they have to prove unit economics. When the hype cycle breaks, people start focusing on the boring work of making AI reliable, maintainable, and genuinely useful for real workflows.

This is when AI actually transforms industries. Not when we're celebrating the next benchmark improvement, but when we're measuring cost savings and productivity gains and workflow transformations that generate actual economic value.

The constraint shifts from "we need more capable models" to "we need to actually deploy the models we have effectively." And that's where all those boring product management and systems engineering skills become force multipliers. The companies that survive won't have the best models—they'll have the best distribution, the clearest use cases, and proven ability to generate profit.

The tech workers who position for this transition—who focus on reliability, integration, and workflow transformation over cutting-edge research—will have 20-year careers in AI infrastructure. The ones chasing the next frontier model will be on a treadmill, jumping from company to company every 18 months as funding and priorities shift.

Let me leave you with a simple framework for evaluating where you stand and what you should do next.

Ask yourself three questions. First, does my value depend on proprietary access to frontier models? If yes, you're vulnerable. You need to shift toward applying commodity models to specific problems. Second, can I articulate my value in terms of business outcomes rather than technical capabilities? If no, start practicing. Learn to speak the language of ROI, efficiency gains, cost reduction, and revenue impact. Third, am I building skills that transfer across the AI hype cycle? Domain expertise, systems thinking, and product judgment will matter long after specific model architectures become obsolete.

The goal isn't to become an AI expert. The goal is to become a domain expert who leverages AI better than anyone else in your field. That's the positioning that creates sustainable careers rather than riding hype cycles.

The soft landing is coming. Hopefully, I’ve given you some things to think about so that you can approach it wisely.

I'd love to hear from you: What domain are you pairing with AI expertise? What positioning strategies are you considering? Hit reply and let me know—I read every response and often turn interesting threads into future newsletters.