Teaching "AI Career Paths" to Architect mode (ages 15–18)

Hook & Warm-Up

No mascot, no soft framing. Read the app's hook line as the opening premise of a strategy discussion:

"The AI career landscape is evolving rapidly — new roles emerge monthly while traditional roles are being redefined. Let's examine career strategy through the lens of comparative advantage."

Open with a harder question than the hook poses on its own: "If new AI-adjacent job titles are appearing faster than any curriculum, including this one, can keep up with, what's actually worth optimizing for in your own preparation — a specific job title, or something more durable?" Let the discomfort sit; there isn't a tidy answer, and the lesson's own claim is that a specific title is the wrong target. The honest answer the lesson builds toward is: optimize for a combination of transferable meta-skills and one area of genuine depth, because the depth compounds in value even as the specific titles around it keep changing.

Worth naming directly, since it's easy to conflate with core AI-technology roles: several of the lesson's twelve example careers (AI policy, AI journalism, AI education, and the four "every field" roles) are not primarily technical roles at all — they're existing professions that now require AI literacy as a component skill. Distinguishing "a job that builds AI" from "a job that requires understanding AI" is the same isAI:true/false distinction the app's own sorting activity uses, and it's the right frame for the whole lesson.

One more framing question worth putting to the class before diving into the scenes: "Every AI capability that exists today was, at some point, a research paper nobody outside a lab had heard of. What does that imply about trying to plan a career around today's specific tools, versus planning around the underlying skills that let you adapt when the tools inevitably change?" Don't resolve this yet — it's the thread the rest of the lesson pulls on.

Main Activity

Treat the three scenes as case studies for two economic ideas — comparative advantage and portfolio-based signaling — rather than as a list of job descriptions to read aloud.

Comparative advantage, applied to careers: The economic concept, originally formulated by David Ricardo to explain why nations trade even when one country is more productive at everything, states that value comes from specializing in whatever a person (or country) can do at the lowest relative opportunity cost — not necessarily the thing they're best at in absolute terms. Applied to a career: your most valuable move usually isn't becoming the single best AI engineer in a room full of AI engineers. It's occupying a position where your specific combination of skills is rare — say, someone with real depth in agriculture, medicine, or music who also has working AI literacy, competing in a much smaller field than "AI engineers" alone. Scene 1's four technical roles (ML engineer, data scientist, AI safety researcher, robotics engineer) are the "absolute advantage in AI" path — genuinely valuable, but also the most crowded and most quickly commoditized as the tools to do this work get more accessible. Scenes 2 and 3's eight roles are all comparative-advantage plays: AI + a domain, where the domain expertise is what's actually scarce.

AI safety, precisely: This age band's quiz uses the term alignment researcher directly, so define it accurately: alignment research is the work of ensuring an AI system's behavior matches its designers' and users' actual intentions — a distinct problem from making a system more capable. A model can become more capable (better at the task it was trained on) without becoming better aligned (more reliably doing what its operators actually want, including in situations its training didn't anticipate). This is why alignment researcher is a growing, dedicated role rather than a subset of general ML engineering — it requires a different set of concerns (robustness, unintended behavior, value specification) than building or scaling a model does. Be precise with students about uncertainty here: current AI safety work addresses documented, present-day problems (bias, factual errors, systems behaving unpredictably outside their training distribution) with reasonable confidence; broader claims about long-term "existential" risk from much more capable future systems are a genuine, actively debated area of research, not a settled consensus — present it as an open question serious researchers disagree about, not as established fact in either direction.

Portfolio-based signaling: Across the tech industry broadly, and increasingly in AI-adjacent hiring specifically, a demonstrated track record of shipped projects, open-source contributions, or published work functions as a stronger signal to employers than credentials alone — partly because the field changes faster than most formal curricula, and partly because a finished project reveals judgment (what did you choose to build, what tradeoffs did you make) in a way a transcript can't. This doesn't make credentials worthless — they still matter for many paths, including if a student is heading to fields like medicine or law where AI is a secondary skill layered on a licensed profession — but for AI-adjacent technical and creative roles specifically, a portfolio has become at least as important as a degree line.

Comparative advantage, applied to the Philippines specifically: This isn't only an abstract framework — it maps onto two real, current pathways available to Filipino students right now. The first is the BPO-to-AI transition: the Philippines is one of the world's largest business-process-outsourcing hubs, and companies operating here (Accenture, Concentrix, Teleperformance, TaskUs, IBM) are shifting roles from pure voice customer service toward AI-assisted quality monitoring, data annotation and labeling, and analytics — work that rewards someone who combines existing BPO-industry experience or strong communication skills with AI literacy, which is a genuinely different, often more accessible, comparative advantage than competing head-on as a generalist ML engineer. The second is remote work for employers based outside the Philippines: a Filipino developer or data scientist strong enough to compete for a remote role with an international client is applying comparative advantage geographically — capturing pay closer to that client's market rate while living and working from the Philippines, rather than competing purely within the local job market. Both are worth naming as concrete, current options, not hypothetical future ones.

Where formal study fits into the portfolio-vs-credentials question: For students weighing a Computer Science or Data Science degree against a self-taught, portfolio-first route, name the concrete Philippine options rather than leaving it abstract: the University of the Philippines, Ateneo de Manila University, De La Salle University, and Mapúa University all offer established Computer Science, IT, or Data Science programs. A degree from one of these remains valuable — for networking, foundational rigor, and fields like medicine or law where AI literacy is a secondary skill layered on a licensed profession — but per the portfolio-signaling point above, it's no longer the only credible route into an AI-adjacent technical or creative career, particularly in fast-moving areas like data annotation, applied ML, or remote development work.

Why "T-shaped" and "meta-skills" both matter here: Given the field's actual rate of change — new tools, techniques, and job titles appearing on a timescale of months rather than years — the most durable preparation is not memorizing today's specific tools, but developing what the lesson calls "adaptable meta-skills": the ability to learn a new tool quickly, evaluate a new claim critically, and update a mental model when the evidence changes. Paired with genuine depth in one domain (medicine, policy, art, whatever a student actually cares about), this combination is what holds up as the specific tools and titles around it keep shifting.

A concrete way to make this less abstract: ask students to name a specific AI tool or technique that was considered cutting-edge five years ago and is now either standard, obsolete, or replaced by something notably better. Most will have some awareness that the tools underpinning today's chatbots and image generators looked very different even a few years back. The point isn't to be alarmed by that pace of change — it's to notice that anyone whose entire career strategy was "master this one specific tool" would have had to restart constantly, while anyone who built strong fundamentals and a habit of fast learning adapted each time with much less friction.

Discussion

• "Comparative advantage says value comes from your relative rarity, not your absolute skill level. Can you think of a case where someone with weaker AI skills than a specialist ML engineer might still be more valuable to an employer? What would make that true?"
• "Alignment research is described as distinct from capability research. Can you explain, in your own words, why a more capable model isn't automatically a better-aligned one?"
• "If AI-adjacent job titles are appearing and disappearing on a monthly timescale, is it still rational to specialize deeply in anything right now, or does that make broad adaptability strictly better than depth? Defend your answer."
• "Portfolio-based hiring rewards demonstrated projects over credentials. What might this system get wrong — whose talent might it undervalue, compared to a credential-based system?"
• "The lesson distinguishes 'practical' AI safety concerns from more speculative long-term ones. How should a serious researcher — or a serious student choosing a career — reason about a risk that's genuinely uncertain, rather than either dismissing it or treating it as settled?"
• "If a specific AI tool you might build a career around today is likely to be obsolete or radically changed within five years, what does that suggest about how much weight to put on any single class, certification, or bootcamp that teaches only that tool?"
• "Between the BPO-to-AI transition and remote work for foreign employers, which better fits your own comparative advantage — and why? Is there a way to pursue both at once?"

Quiz Walkthrough

Wrap-Up & Extension

Close with: "Every career strategy in this lesson — comparative advantage, portfolio signaling, T-shaped skill combinations, adaptable meta-skills — is a response to the same underlying fact: this field is changing faster than any institution, including this app, can fully keep up with. That's not a reason to panic or to disengage from planning. It's a reason to invest in the parts of your preparation that don't expire — genuine depth in something you care about, the habit of learning quickly, and a body of real work that proves both."

Extension activity — Comparative Advantage Case Study: Have students individually map their own potential comparative advantage: list one domain they have real depth or genuine interest in (not necessarily AI-related — could be agriculture, health, law, art, sports, language), then research one specific way AI is currently being applied in that domain, and identify what unique value someone with both that domain depth and AI literacy could offer that a pure AI generalist couldn't. Students who don't have an obvious domain in mind can research the BPO-to-AI transition or remote work for foreign employers instead — both are real, current, and directly relevant to the Philippine job market they'll actually enter. Have each student write a short paragraph and, if time allows, present it to the class. This extends the 10–15 minute core lesson into a full period and turns an abstract economic concept into a genuinely personal, actionable career hypothesis.

A useful stretch for the strongest students: have them also identify one weakness in their own comparative-advantage argument — a reason a pure AI generalist, or someone with a different domain, might actually be better positioned than they are. Comparative advantage as a framework can become a comforting story that flatters whatever a student already wants to do; making them argue against their own case is a genuine test of whether they understand the concept or are just using it to justify a decision they'd already made.