Student Action Guide 2026 — May Update

What Students
Should Do in the AI Era

Student Action Guide for the AI Era
May 2026. Stanford AI Index 2026 confirms it: software engineering employment for 22-25-year-olds is down 20% from 2024.
Yet ZipRecruiter reports that job-seekers who use AI heavily land offers at a 76% rate (vs 33% for non-users).
The key is becoming the one who uses AI — not the one replaced by it. A current guide for middle, high, and university students.

-20%
SWE employment decline,
ages 22-25 vs 2024 (Stanford AI Index 2026)
76%
Offer rate for heavy AI users
(33% for non-users)
(ZipRecruiter Q1 2026)
5.6%
Unemployment, ages 22-27
(overall rate: 4.2%)
(NY Fed 2026)
+56%
Wage premium for
workers with AI skills (PwC)
↓ scroll
Reality Check

Face the Reality

The world today's students are stepping into — what the data says about both the opportunity and the threat.

⚠ The Entry-Level Job Crisis

Entry-level postings are collapsing

-35%

Entry-level job postings are down 35% globally. UK tech graduate hiring fell 46% in 2024, with another 53% drop forecast for 2026.
— WEF / IntuitionLabs 2025-2026

New graduates without jobs

33%

33% of 2025 graduates were unemployed and still job-hunting (up sharply from 20% in 2024). Only 30% landed roles related to their degree.
— Cengage 2025

White-collar entry roles disappearing — the ServiceNow CEO's warning

30-35%

ServiceNow CEO Bill McDermott (March): "Graduate unemployment could climb into the 30-35% range within two years." Dario Amodei: "Within one to five years, half of entry-level white-collar jobs could disappear; software engineers can be replaced in 6-12 months." Anthropic Economic Index: programmer roles top observed AI exposure at 74.5%.
— CNBC / Fortune / Anthropic 2026

Stanford AI Index 2026: only the young get cut

-20%

Stanford AI Index 2026 (released April 13): SWE employment for 22-25-year-olds is down 20% from 2024. Senior engineers in the same year actually grew — the asymmetry "AI replaces the young, complements the experienced" is now established. NY Fed: 22-27 unemployment is 5.6% vs 4.2% overall. Goldman: AI is removing roughly 16,000 US jobs per month, concentrated on Gen Z. GenAI's adoption curve (53% of the population in 3 years) is the fastest in human history — faster than the PC, faster than the internet.
— Stanford HAI / NY Fed / Goldman 2026

The mass exodus from CS majors

-11.2%

For US CS majors in academic year 2025-2026: the steepest decline of any field at -11.2% (computer & information sciences overall down 8.1%, National Student Clearinghouse). In their place, mechanical engineering is up 11% and electrical engineering up 14%. Roughly half of US college students are now considering changing majors because of AI (early 2026 survey). Northwestern, Columbia, and USC all launch dedicated AI programs in Fall 2026, accelerating the redistribution from CS toward AI / data science.
— Washington Post 2026/4 / TechCrunch 2026/2

University of Tokyo entrance exam: top AI models score in the admit zone

April 30, 2026: a study fed the actual February 2026 University of Tokyo entrance exam to ChatGPT 5.2 Thinking, Gemini 3 Pro Preview, and Claude Opus 4.5. If the AIs had taken the test, they would have ranked at the top of the admitted class. Japan's Common Test for University Admissions (共通テスト) was sat by 496,000 students in January 2026 — the line between "AI is fine here" and "AI is forbidden here" is blurring fast.
— Japan Times 2026/4/30

📊 How students are actually using AI

Students using AI — effectively 100%

95%

HEPI 2026 survey of 1,054 UK undergraduates (December 2025): 95% use AI in some form, 94% use it for coursework. What was a novelty three years ago is now near-universal. 68% say AI skills will be essential for their future, but only 48% feel supported by their instructors. 36% sense their school is signaling "don't use it" — the gap between students and educators is widening.
— HEPI Report 199 (released March 2026)

High schoolers and homework

84%

84% of high school students use AI for homework. 66% recognize that "leaning on it too much will dull my thinking."
— College Board 2025

Teen chatbot use

66%

About 66% of US teens use AI chatbots; 30% use them daily. Parents underestimate this by 13 percentage points.
— Pew Research 2025

The AI-dependence paradox

OECD research: students using AI tools improved task success by 48%, but their performance without AI dropped 17%. 86% of college students now use AI as their primary research tool.
— OECD 2026 / Harvard Education 2026

AI has become the de facto hiring filter

76% vs 33%

ZipRecruiter Q1 2026: job-seekers who use AI heavily get offers at 76%, vs 33% for those who don't — more than double. Using AI for schoolwork should be reframed: not "cheating," but "a baseline requirement for getting hired." US universities are moving from blanket bans to graduated guidelines (Inside Higher Ed, Feb 2026).
— ZipRecruiter / Inside Higher Ed 2026

Expert Voices

Expert Voices

Messages from the leaders shaping AI education.

SK
Sal Khan
Founder, Khan Academy
AI Education Revolution

Khanmigo reached 2 million users in the 2024-25 school year (up 731% YoY) — bringing a Socratic AI tutor into classrooms worldwide. Khan Academy itself runs at 17.7M monthly learners. 380+ US districts have adopted it.

"AI being used to amplify human intelligence (HI) — that's the most powerful and most poetic use case of all."
AS
Andreas Schleicher
Director for Education and Skills, OECD
Learning Compass

Architect of the OECD Learning Compass 2030. His thesis: the era when "learning was your job" is over — now learning itself is the job.

"The future of work is deeply human — rooted in human consciousness, the ability to navigate complex relationships, and ethical judgment."
MR
Mitchel Resnick
Professor, MIT Media Lab
Creative Learning

Creator of Scratch. Champions creative learning through the "4 Ps" framework — Projects, Passion, Peers, Play. Warns that AI should support learners, not dominate them.

"The disruption from AI is forcing us to rethink education — and that's a chance to push it in the direction we actually want."
KL
Kai-Fu Lee
CEO, Sinovation Ventures
Curiosity, Critical Thinking, Creativity

Argues that "the work that survives requires three human skills: curiosity, critical thinking, and creativity." Frames AI's "Great Reshuffle" as a redefinition of jobs, not a wholesale loss of them.

"AI's impact isn't job loss — it's job redefinition. Compete on what only humans can do."
DA
Dario Amodei
CEO, Anthropic
Warning on entry-level jobs (most relevant to students)

Warns that "within 1-5 years, half of entry-level white-collar jobs could disappear," and that software engineers may be replaceable in 6-12 months. Puts his p(doom) at 25%. In April, Anthropic surpassed OpenAI on revenue at $30B ARR, and published the measured AI exposure for programmer roles at 74.5%.

"An AI smarter than a Nobel laureate could arrive as soon as 2027. Half of entry-level jobs could be gone in 1-5 years — policymakers need to look at this directly."
SA
Sam Altman
CEO, OpenAI
From AI Tool Mastery to Societal Reform

"Mastering AI tools is the new top-priority skill — the equivalent of 'learn to code' a generation ago." April 2026: published the 13-page "Industrial Policy for the Intelligence Age" (shifting tax from labor to capital, four-day workweek) — followed four days later by a Molotov attack on his home. April 23: officially released GPT-5.5 "Spud", scoring 82.7% on Terminal-Bench 2.0 and retaking the industry top spot.

"By the end of 2028, the intelligence inside data centers will exceed that of every human on the outside, combined."
— April 2026, "New Deal for the AI Era"
DH
Demis Hassabis
CEO, DeepMind / 2024 Nobel Prize in Chemistry
Direct Advice to Students

Won the 2024 Nobel Prize in Chemistry for AlphaFold. At the Davos 2026 dialogue with Amodei (1-2 years), Hassabis took the more measured "5-10 years" position on AGI, putting AGI by 2030 at roughly 50% probability. Stanford AI Index 2026: Gemini Deep Think won gold at the International Mathematical Olympiad. Hassabis: "In verifiable domains like code and math, AI is moving very fast — but scientific discovery and creative reasoning are still hard."

"If I could give an undergrad one piece of advice — get scarily good with AI tools. Everything else flows from there."
— Davos 2026
BM
Bill McDermott
CEO, ServiceNow
The 30-35% Graduate Unemployment Warning

On CNBC in March 2026: "Graduate unemployment is currently 9%, but it could climb into the low 30s within two years." Predicts 3 billion non-human digital agents will be added to enterprises by 2030. Gen Z is concentrated in routine white-collar work — data entry, customer support, paralegal tasks, billing — which makes them the most exposed to AI.

"This isn't a warning, it's a forecast. Only the students who prepare get the future."
— March 2026, CNBC
Future-Proof Skills

Skills AI Can't Replace

The core skills WEF Future of Jobs 2025 says will matter most by 2030 — 66% of leaders say they "won't hire without AI literacy." Workers with AI-related skills earn a 56% wage premium (PwC). The share of jobs requiring a degree dropped from 66% to 59% — hiring is shifting toward skills-based.

🧠

Analytical Thinking + AI Evaluation

Analytical Thinking + AI Evaluation

70% of employers rank this as the single most important skill. The 2026 job-title equivalent: "Evaluation Engineer" instincts. The ability to critically judge AI output and structure ambiguous problems — Mollick's "Jagged Frontier" awareness in action.

🎓 From middle school: scientific hypothesis testing, debate, the habit of fact-checking AI output.
💪

Resilience & Agility

Resilience & Agility

The pace of skill change is accelerating 66% over the next five years. Adapting under shifting conditions and learning from failure is now a career-survival skill.

⚽ From middle school: team sports, taking on tough goals.
🎨

Creative Thinking

Creative Thinking

AI can recombine existing patterns, but it can't generate "lived intent or empathic structure." Ranked #4 on the WEF skills list.

🎵 From middle school: music, visual art, theater, project-based learning.
🔍

Curiosity & Lifelong Learning

Curiosity & Lifelong Learning

39% of all skills will shift by 2030. As OECD puts it: "The era when learning was your job is over — now learning itself is the job."

📖 From middle school: deep reading habits, inquiry-based projects.
🤝

Empathy & Active Listening

Empathy & Active Listening

AI can simulate empathy, but it can't experience emotion. McKinsey forecasts demand for social and emotional skills will rise 14% by 2030.

👥 From high school: mentoring, volunteering, cross-cultural exchange.

Ethical Reasoning

Ethical Reasoning

"AI models have no truth and no judgment of their own — humans have to impose the parameters" (McKinsey CEO). Filling that gap is a uniquely human role.

💬 From high school: philosophy, AI ethics, bias detection.
💻

AI Fluency

AI Fluency

One layer above "AI literacy." Anthropic's AI Fluency course defines it as "the ability to work with AI effectively, efficiently, ethically, and safely — while understanding its limits." US demand for AI-fluent workers grew 7x in two years. The US Department of Labor published a national framework on Feb 13, 2026. Vocabulary: hallucination, sycophancy, automation bias, jagged frontier.

🚀 From middle school: Scratch → Python → AI collaboration → eval design.
🔍

Oversight / Centaur Mode

Oversight / Centaur Mode

Harvard 2026's three-way split — Cyborg 60% / Centaur 14% / Self-Automator 27% — found Centaurs achieved the highest accuracy. Don't dump everything on AI (that's the Self-Automator "no-skilling" trap); use it through "Directed Knowledge Co-Creation," staying in command. With 88% of agents failing in production, companies are actively recruiting "AI Reliability Engineer (ARE)" candidates.

🧑‍🔧 From high school: verifying AI output, the basics of Spec-Driven Development.
🎯

Leadership & Social Influence

Leadership & Social Influence

Ranked #3 on the WEF skills list. Maximizing a team's output and aligning stakeholders is something AI can't substitute for.

🏆 From high school: leading a club, student council, founding something.
Warning

What NOT to Do

The traps students fall into in the AI era — and the right move instead.

⚠ The cognitive offloading risk — latest 2026 research

If you outsource "thinking" to AI, your cognitive muscles atrophy. Oxford research: students who lean heavily on AI for homework score lower on problem-solving tests. RAND survey: 60% of students worry that AI use is eroding their critical thinking. Microsoft / CHI study: the more you trust AI, the less critical thinking you do. International AI Safety Report 2026: "Models are starting to find loopholes in evaluations and hack their reward signals" — the risk that students swallow AI's shallow pattern-matching answers is growing.

❌ Hand the answer over to AI

Get AI to do your homework wholesale. Copy-paste your essay. Skip the "thinking" step entirely.

✔ Think alongside AI

Reason it out yourself first, then use AI as a sparring partner. Have it poke holes in your reasoning. Use Socratic dialogue to go deeper.

❌ Over-specialize too early

Believing "only AI engineers will survive" and pouring everything into AI/CS from middle school onward. CS enrollment is already down 8.1%.

✔ M-shaped breadth

Build more than one "peak." McKinsey, Microsoft, and Anthropic are all re-valuing humanities backgrounds. The intersection of "technology and the humanities" is the strongest place to stand.

❌ Dismiss the humanities and liberal arts

"With AI, who needs the humanities?" is a dangerous misread. English enrollment is dropping at -5% CAGR, history at -3% CAGR.

✔ Fuse STEM with the liberal arts

Microsoft Chief Scientist: "Meta-cognitive skills — flexibility, adaptability, critical thinking — depend on a liberal arts education."

❌ Abandon deep reading

Lean on AI summaries, fall into a "TL;DR mindset," and lose the sustained attention required to read long-form text.

✔ Read deeply and use AI

Read complex texts on your own and build mental schemas. Treat AI as something that amplifies your reading, not replaces it.

❌ Sideline the body and the real world

More screen time correlates positively with adolescent depression and anxiety. Digital tools don't substitute for embodied learning.

✔ Digital detox plus physical activity

One week of social media detox: anxiety -16.1%, depression -24.8%. Team sports build communication, coordination, and resilience.

❌ Aim to become a "Prompt Engineer"

The "Prompt Engineer" role hyped in 2022-24 isn't a standalone job anymore. As models advanced, the "magic-incantation" skill rapidly commoditized. Karpathy: "Context Engineering matters more than prompt engineering."

✔ Learn Context / Harness Engineering

The real craft is designing what you give the AI and what environment you let it work in: information architecture, evaluation design, agent operations. By late 2026 this is evolving into "Harness Engineering" — the higher-order layer that designs the AI's entire working environment.

❌ Don't naively believe Human + AI is always stronger than AI alone

In chess this premise has already collapsed as of 2026. If a human overrides Stockfish, they almost always make it worse. Advanced Chess tournaments have effectively died. Believing "I'm human, therefore I add value" without conditions is dangerous.

✔ Audit "where humans still win" every quarter

The boundary moves quarter to quarter. In knowledge work, humans still hold an edge in ambiguity, ethics, and contextual integration — but that list isn't fixed. Build a habit of periodically re-evaluating the AI exposure of your target field with measured data like the Anthropic Economic Index.

Action Plan

Age-Specific Action Plan

A concrete to-do list for your stage of education.

AI & Technical

Understand "what's inside AI" from first principles

  • Visual programming with Scratch or MIT App Inventor
  • MIT Day of AI (last Friday of March every year, free) — 2M+ students across 50 states and 175 countries take part
  • The Code.org AI course (now folded into CS Discoveries, free) — move from "AI user" to "AI builder"
  • Before asking AI a question, build the habit of asking: "Why would the answer come back this way?"
  • With family or friends, practice splitting tasks into "OK to delegate to AI" vs. "must do myself"
Human Skills

Build the human core AI can't take

  • Team sports — develop cooperation and resilience
  • Music, visual art, theater for creativity — own your intent and your story, not just the surface forms AI can imitate
  • Deep reading habit — finish at least one full-length book a week (no escaping into AI summaries)
  • Volunteering and community work — train empathy in the flesh, not on a screen
  • At least one hour a day away from screens — APA's recommendation for AI-native generations' mental health
Academic

Learn how to learn

  • Build broad foundations across math, science, and the humanities — the base for an M-shaped, multi-peak profile
  • Growth mindset — replace "I can't" with "I can't yet"
  • Engage in scientific experimentation and hypothesis testing
  • Stay aware of the three skill types in the OECD Learning Compass 2030: cognitive, social-emotional, and physical
  • Practice "asking better questions" — the quality of the question, not the answer, is the differentiator
Extracurricular

Make your "first AI experience" outside the classroom

  • Debate and speech for logical expression (the kind of thinking that lets you push back on AI)
  • Try the USAII Global AI Hackathon (June every year, free, 12-18 track available, $15K prize pool)
  • Share Day of AI projects in your school paper or on social media
  • Cross-cultural exchange — connect with peers abroad. Now that AI translation flattens the language barrier, the cultural barrier is the one worth practicing on
  • Keep a "collection of AI's wrong answers" — let yourself feel the Jagged Frontier in your bones
AI & Technical

Build AI Fluency (not Prompt Engineering)

  • Python plus Joho I (Information I) — a required Common Test subject from 2025, 69.3 average — for text-based programming
  • Anthropic AI Fluency course (free, on Skilljar) — learn hallucination, automation bias, and Jagged Frontier
  • Use Claude Code / Cursor / GitHub Copilot on real projects to feel the line between "what I write" and "what AI writes"
  • Entry point to Spec-Driven Development — practice writing what you want to build as a spec (the foundation for ARE work later)
  • Train yourself to spot AI-generated misinformation and deepfakes (Hinton, May 2026: "AI has gotten smarter — at deception too")
Human Skills

Train yourself to not become a Self-Automator

  • Lead in clubs and sports — practice leadership and teamwork
  • Mentor younger students to grow your empathy
  • Digital wellness — try a one-week screen detox (evidence: -16.1% anxiety, -24.8% depression)
  • Deliberately raise your emotional intelligence — Microsoft's "5C": Critical thinking, Curiosity, Creativity, Communication, Collaboration
  • Once a month, self-test: "could I produce the same quality without AI?" — the share of students slipping into Self-Automator mode (27%) is rising
Academic

Reach into AI-resistant territory

  • STEAM — STEM crossed with the arts. Microsoft Chief Scientist: "Meta-cognition is impossible without the liberal arts."
  • Deep reading of complex texts — philosophy, history, literature. Don't escape into AI summaries.
  • Strengthen research skills and scientific inquiry (now reflected in the Common Test)
  • Foundations of philosophy and ethical reasoning — get good enough to discuss AI ethics in your own words
  • Foreign language study to widen your M-shape (the goal is to learn the limits of AI translation)
Extracurricular

Your portfolio should be "things that run"

  • USAII Global AI Hackathon 2026 (June 14-21, $15K prize pool) / USAII AI NextGen Challenge (June 20-21 in Atlanta, $100K total)
  • MIT America's Youth AI Festival (every July, winners get a free invitation)
  • DECA / Junior Achievement for entrepreneurial experience / Model UN / debate
  • Study abroad and overseas exchange — now that AI translation lowers the language load, the value of cultural experience is rising
  • Three or more "running AI products" on your personal GitHub — admissions officers and recruiters will read this as "ARE candidate"
  • Devpost runs AI hackathons year-round (many are online and free)
AI & Technical

The "AI Reliability Engineer" roadmap

  • Your major × AI (law×AI, medicine×AI, art×AI, etc.) — Stanford AI Index: pure programmer roles sit at 74.5% exposure; domain fusion is the defensive line
  • Spec-Driven Development (DeepLearning.AI short course) plus Claude Code / Cursor — train yourself to ship production-grade code
  • Master Eval (evaluation engineering) — LangSmith / Promptfoo / Braintrust. The skill seniors say juniors most lack.
  • Land production PRs in open-source AI projects (MCP, Anthropic SDK, LangChain, etc.) — they're the receipt you bring to interviews
  • Build hallucination-checking and reward-hacking detection into real projects (per the International AI Safety Report 2026)
Human Skills

Aim for the "Directed Centaur 14%"

  • Treat AI not as a "collaborative partner" but as a "hyper-capable new hire who needs supervision"
  • Harvard 2026: 14% Centaurs hit peak accuracy / 27% Self-Automators fall into the "no-skilling" trap
  • When AI confidently gets it wrong, log it that week — build your personal map of the Jagged Frontier
  • Update your Cyborg / Centaur / Self-Automator self-diagnosis every quarter
  • AI's weak points: ambiguity, ethics, real-time contextual reading — invest in face-to-face collaboration there
  • Deep mentorship (both receiving and giving) — tacit knowledge is the last thing AI absorbs, and your last clean differentiator
Academic

M-shaped cross-domain expertise

  • Liberal arts × technical (McKinsey 2026: actively recruiting "humanities-trained people with creativity and judgment")
  • Map your curriculum onto Purdue's five domains: Learning with AI / about AI / Researching AI / Using AI / Partnering in AI
  • If you're a pure CS major, don't take the path on faith — US CS enrollment is down 11.2%. Consider applied AI / data science / Bio×AI / robotics / computational neuroscience blends instead
  • Independent research projects — and the habit of critically evaluating AI output
  • Be fluent in both design thinking and Spec-Driven methodology
Career Preparation

Hands-on practice for the "ARE hiring track"

  • Anthropic Fellows Program (May and July 2026 cohorts, $3,850/week + $15K/month compute)
  • OpenAI Residency / Allen AI Internship / Google AI Residency
  • ZipRecruiter Q1 2026: heavy AI users get offers at 76% vs 33% — bring concrete AI-collaboration stories to your resume and interviews
  • One production-running agent on your personal GitHub (with 88% failing in production, just working is rare)
  • Anthropic's Code with Claude (SF May 6 / London May 19 / Tokyo June 10) / Devpost / Kaggle — get inside real practitioner communities
  • Founding experience and startup internships — Stargate $100B, Anthropic $40B ARR, Salesforce Agentforce $540M ARR. The capital is there.
  • Search for new job titles: AI Editor, Forward-Deployed Engineer, Context Engineer
AI × Learning

The Right Way to Learn with AI

From "ask AI for the answer" to "think with AI."

❌ Wrong patterns

Have AI write the report and copy-paste it
Memorize the answer it gave you
Trust AI output without checking
Read only the summary, never the source
Ask AI before thinking yourself

✔ Right patterns

Reason it through yourself, then sanity-check with AI
Ask AI to quiz you to confirm you actually understand
Have AI find the flaws in your reasoning
Use Socratic dialogue to reach deeper understanding
Always verify AI output against the original source

💡 AI education programs to watch

Millions of students are already learning to use AI the right way. Harvard RCT: AI-tutored students gained 0.73-1.3 SD effect size — well above classroom learning. Purdue is the first US university to roll out an AI graduation requirement, applied to all 44,000+ undergraduates from Fall 2026 — across five domains: Learning with AI / Learning about AI / Researching AI / Using AI / Partnering in AI. Through a $35M partnership with Google, Purdue is deploying AI software to every student and staff member.

5 domains
Purdue AI graduation requirement
First in the US (from Fall 2026)
2M
Khanmigo users
2024-25 school year (+731% YoY)
2M+
Students reached by MIT RAISE
50 states + 175 countries
$1B
Google education investment
10M+ students over 3 years
20K+
Anthropic + CodePath
university student program

🌐 OECD × European Commission AI literacy assessment is here

In May 2025, OECD and the EC jointly released a draft AI Literacy Framework for Primary & Secondary Education; the official version lands in 2026. PISA 2029 Media and AI Literacy (MAIL) will assess 15-year-olds worldwide on active, critical AI use (results expected December 2031). In PISA 2025, "Learning in the Digital World (LDW)" was added as a new domain — results scheduled for September 8, 2026. AI literacy has been formally promoted from "optional" to "internationally assessed."

Global Map

The Global AI Education Map

UNESCO: only 11 countries have a developed K-12 AI curriculum.

🇯🇵

Japan

Active

Joho I (Information I) has been required for all high school students since 2022. The first Common Test for University Admissions (共通テスト) including it ran in January 2025: 279,000 takers, average 69.3. Through GIGA School every K-12 student has a 1:1 device. MEXT survey: 58% of teachers feel uneasy teaching AI (Education Accelerator has trained roughly 50,000 of them). The AI Promotion Act took effect in May 2025; the AI Basic Plan was approved by Cabinet in December. From May 2026 to March 2027: the government AI "Gennai" rolls out to 180,000 staff across all ministries (under the Takaichi administration, led by the Digital Agency). April 2026: Microsoft commits $10B (¥1.6T) to Japan, with a goal of training one million engineers by 2030. Comprehensive partnership with Keio University. By 2040: a projected shortage of 3.26M AI / robotics workers.

🇨🇳

China

Mandatory

From September 2025, AI education is mandatory in all primary and secondary schools. Beijing alone: 1,400+ schools at minimum 8 hours per year. Tiered curriculum: elementary (basic concepts) → middle (applications) → high school (innovation). Target full rollout by 2030, textbook integration by 2035.

🇸🇬

Singapore

Active

Since 2025, every primary and secondary school runs the AI for Fun module (5-10 hours). AI is woven into the national Code for Fun program. The Smart Nation Educator Scholarship is producing 250 specialist teachers.

🇫🇮

Finland

Integrated

Elements of AI has been completed by over 1 million people from 110+ countries. In 2025, Finland integrated AI literacy from early childhood through vocational training. About 2% of Finland's population has finished the course — the highest rate in the world.

🇺🇸

United States

Partial / State-Level

60% of public high schools offer foundational CS courses. April 2025: the executive order "AI Education for American Youth" was signed. Ohio: first state to require an AI policy in every K-12 district. 2026: 134 AI-education bills introduced across 31 states (MultiState) — California AB 1159 (no training AI on student data), Idaho SB 1227 (AI tools must protect personal information), Oklahoma and Maryland (AI banned for high-stakes decisions). Purdue introduces the first US AI graduation requirement (5 domains) for all 44,000 undergraduates from Fall 2026.

🇬🇧

United Kingdom

In Development

The national curriculum doesn't explicitly cover AI yet. The Raspberry Pi Foundation is leading the research on AI curriculum. A new data science and AI qualification for 16-18-year-olds is in proposal. AI Education Symposium held February 2025.

Framework

OECD Learning Compass 2030

"Whatever is easy to teach and easy to test is also easy to digitize and automate." — OECD

Student Agency

The capacity for students to set their own direction with purpose and responsibility — the heart of the Learning Compass.

Creating New Value

Creating New Value

Combining existing knowledge to produce innovative ideas and products. Built on curiosity, imagination, and critical thinking.

Reconciling Tensions & Dilemmas

Reconciling Tensions & Dilemmas

Holding contradictory demands and viewpoints in balance. Navigating trade-offs like fairness vs. freedom, individual vs. society.

🎯

Taking Responsibility

Taking Responsibility

Owning the outcomes of your actions and your learning. The base layer of ethical judgment and ethical action.

🔄

Anticipation-Action-Reflection

Anticipation-Action-Reflection

Anticipate the future, act, then reflect on the result. The engine that compounds your learning over time.

🧠

Cognitive & Meta-cognitive

Cognitive & Meta-cognitive

Analytical thinking, creative thinking, and learning how to learn. The first of three skill categories.

Social & Emotional

Social & Emotional

Empathy, collaboration, resilience. The capacity to engage with others and contribute to society — the largest gap between humans and AI.

Resources

Resources You Can Start Today

World-class AI education programs, free at any age.

Khanmigo

Khan Academy

A Socratic AI tutor that won't give you the answer — it makes you think. Reached 700,000 K-12 students / 2M cumulative users in 2024-25, adopted by 380+ US districts. 1M+ projected for 2025-26.

Middle High University

MIT Day of AI

MIT RAISE

Reaches 2M+ students across 50 states and 175 countries. Free curriculum and teacher training. AI Festival scheduled for July 2026.

Middle High University

Code.org / Scratch

Code.org / MIT

Free K-12 coding curricula. Scratch is the world's largest programming platform for children. Both have added AI learning modules.

Middle High University

Google AI for Education

Google

$1B investment. 100+ public universities are part of the AI for Education Accelerator. Free Gemini training for 6M US teachers. AI Quests delivers game-based learning to middle schoolers.

Middle High University

Elements of AI

University of Helsinki + Reaktor

A free online course completed by over 1M people in 110+ countries. AI fundamentals explained for non-engineers. About 40% of completers are women.

Middle High University

fast.ai / CS50

fast.ai / Harvard

fast.ai: a hands-on deep learning course (free). CS50: Harvard's legendary CS intro, free on edX.

Middle High University

AI4ALL

AI4ALL

A diversity-focused summer AI program for high schoolers (held at Stanford and other universities; free or scholarship-based). The Ignite career accelerator serves university students.

Middle High University

Claude for Education

Anthropic

Includes a Socratic learning mode. Adopted by Northeastern University (50K users) and others. In partnership with CodePath, providing AI education to 20,000+ students.

Middle High University

Anthropic AI Fluency Course

Anthropic / Skilljar

The official course for the "AI Fluency" concept that became standard in 2026 (free). Teaches the vocabulary — hallucination, automation bias, Jagged Frontier, O-ring automation — systematically.

Middle High University

Anthropic Fellows Program

Anthropic

May and July 2026 cohorts. $3,850/week + $15K/month compute, four months of AI Safety research. Frontier topics: Scalable Oversight, Mechanistic Interpretability, Model Welfare. Open to undergraduates.

Middle High University

USAII Global AI Hackathon 2026

USAII

Held every June, open to grades 9-12 and university students. $15K total prizes, with mentoring and office hours. The AI NextGen Challenge (June 20-21 in Atlanta, $100K total) runs alongside.

Middle High University

Day of AI 2026 / Youth AI Festival

MIT RAISE

A worldwide simultaneous event held the last Friday of March (March 27 in 2026). MIT America's Youth AI Festival in July invites top participants to MIT. Reaches 2M+ students across 50 states and 175 countries. Fully free with materials provided.

Middle High University

Spec-Driven Development course

DeepLearning.AI / GitHub Spec-Kit / Kiro

The production-grade AI development methodology that crystallized in 2026. A spec-driven approach designed to push past the "three-month wall" of technical debt that vibe coding hits. Hands-on integration with Claude Code and Cursor is now the norm.

Middle High University

Code with Claude developer conference

Anthropic

Anthropic's first full-scale developer conference, in 2026. SF May 6 → London May 19 → Tokyo June 10. Topics: Agentic Coding, MCP, Constitutional AI, Production Reliability. Virtual attendance is free; recordings are made available.

Middle High University
Start Today

Your First Steps Today

Check items off — your progress is saved.

Completion
0%

🎓 First steps for middle schoolers

Finish one project in Scratch or MIT App Inventor
Build the habit of answering yourself before asking AI (think first → AI → compare)
Read one full book this month (no escaping into AI summaries)
Join a team sport or a group activity
Track your screen time for a week and review it
At Day of AI 2026 (March 27) or in school, learn "why does AI answer this way?"
Talk with your family about which tasks belong to AI and which belong to you
In a week, log one thing AI got wrong (feel the Jagged Frontier)

Sources

WEF — Future of Jobs Report 2025 PwC — Global AI Jobs Barometer 2025 OECD — Learning Compass 2030 Dario Amodei — The Adolescence of Technology (2026) Fortune — ServiceNow CEO: Graduate Unemployment Warning Cengage — 2025 Graduate Employability Report Programs.com — AI Education Statistics 2025 College Board — High School AI Usage Survey 2025 Pew Research — How Teens Use and View AI (2026) Dallas Fed — Young Workers and AI-Exposed Occupations (2026) IntuitionLabs — AI Impact on Graduate Jobs WEF — Gen Z Labour Market and AI Economy Fortune — Liberal Arts in the AI Age (Microsoft) Fortune — McKinsey Hiring Liberal Arts Majors (2026) RAND — Student AI Use and Critical Thinking (2026) Microsoft / CHI 2025 — GenAI and Critical Thinking Oxford — AI Homework and Problem-Solving (2025) Frontiers — AI Education Across School Levels (2024) Khan Academy — Annual Report (Khanmigo) MIT RAISE — Responsible AI for Social Empowerment Google — Education Year in Review 2025 Anthropic — Claude for Education Elements of AI — University of Helsinki UNESCO — K-12 AI Curricula Mapping AI Track — China Mandates AI Education Singapore IMDA — Code for Fun AI Track — Finland AI Education Model White House — AI Education for American Youth (2025) Code.org — State of CS Education 2025 Raspberry Pi Foundation — Data Science & AI Qualification McKinsey — The Skills Revolution Stanford HAI — AI Index Report 2025 APA — Health Advisory: AI and Adolescent Well-being (2025) Frontiers — Digital Detox and Mental Health (2025) TechCrunch — CS Enrollment Decline (2026) NACE — What Employers Look For in Graduates CNBC — AI Agents and Graduate Unemployment MEXT — Generative AI Guidelines v2.0 Dallas Fed — AI Substitutes Young Workers, Complements Experienced (2026) OECD — AI and Student Achievement Paradox (2026) Japan AI Promotion Act — Innovation-First Blueprint (FPF) Sam Altman — "New Deal for the AI Era" (Apr 2026) Purdue University — AI Graduation Requirement (2026) Microsoft — $10B Japan AI Investment (Apr 2026) BCG — AI and the Entry-Level Job Crisis (2026) Fortune — Goldman: AI Cutting 16K US Jobs/Month, Gen Z Brunt (Apr 2026) Anthropic — Labor Market Impacts (Observed Exposure, 2026) CNBC — ServiceNow CEO: Graduate Unemployment Could Hit 30-35% Washington Times — Purdue First US Univ. AI Graduation Requirement OECD — PISA 2029 Media & AI Literacy Assessment AI Literacy Framework — OECD/EC (Final 2026) International AI Safety Report 2026 (Bengio, 100+ experts) Davos 2026 — Amodei vs Hassabis on AGI Timeline & Student Advice inAmericaEdu — Stanford, Purdue, Dartmouth Redefining Learning (2026) Tom's Hardware — Q1 2026: 80K Tech Layoffs, ~50% AI-Driven TechCrunch — Anthropic $30B ARR, Passes OpenAI (Apr 2026) Gallup — AI Routine for College Students (2026) The Centaur's Dilemma: What Chess Teaches About AI Era (2026) MIT Sloan — Cyborg / Centaur / Self-Automator (2026) Ethan Mollick — Jagged Frontier Prompt Engineering Is Dead — Context Engineering (2026) Chess.com — Centaur Approach in AI Era Stanford HAI — AI Index Report 2026 (Apr 13) Stanford AI Index — 12 Takeaways: SWE 22-25 -20% ZipRecruiter Q1 2026 — AI Users 76% vs 33% Hire Rate MultiState — 134 AI Education Bills in 31 States (2026) Inside Higher Ed — Faculty Moving Away from AI Bans (2026) Japan Digital Agency — Government AI Gennai 180K-staff Pilot OpenAI — GPT-5.5 "Spud" (Apr 23, 2026) Anthropic — Code with Claude Conference (May 6 SF)