DL 416

Catastrophic Forgetting

Published: December 14, 2025 • 📧 Newsletter

Conversations about children and technology have moved beyond “access vs. safety.” The real question now is: how do we prepare youth to navigate, contribute to, and shape digital spaces responsibly, ethically, and safely?

Yet while we debate pedagogy, the systems that make digital learning possible are failing. Energy grids strain under the weight of AI data centers. School databases leak millions of student records. Teachers often lack the training, resources, and infrastructure to deliver the digital literacy we claim to prioritize.

Policy matters. But infrastructure determines what is actually possible. And right now, both policy and infrastructure are showing cracks. Creating a kind of systemic “catastrophic forgetting” that risks leaving a generation unprepared.

This will be my last issue for 2025. I'll see you in the new year.

If you’ve found value in these issues, subscribe here or support me here on Ko-fi.

🔖 Key Takeaways

• Innovation Outpaces Oversight: Federal AI policy accelerates platform deployment while weakening state-level protections, leaving schools and students exposed to untested systems.
• Complexity Overload in Education: Fragmented federal funding structures force districts and states to navigate multiple agencies, illustrating the gap between policy design and operational reality.
• Technical Limits Matter: AI leaders at Google, OpenAI, and Anthropic highlight that current models face practical ceilings, driving research into continual learning, but “catastrophic forgetting” shows progress is not guaranteed.
• Global Infrastructure Strategies Diverge: While the U.S. bets big on frontier AI, China invests in manufacturing, energy, and batteries. Reminding us that broader infrastructure choices shape long-term digital futures.
• Private Data Centers vs. Public Systems: Rapid investment in AI infrastructure is straining labor and resources, delaying public works and revealing how physical systems constrain education and societal resilience.

📚 Recent Work

I’m continuing my work on digital sovereignty, trust, and agency:

• You Downloaded Signal. Now What? — A practical starting point for reclaiming control over digital communication.
• What Does It Mean to Be a Live Human Signpost? — On guidance, responsibility, and human presence in moments of uncertainty.

🧭 A Week of Policy Contradictions

This week marked a quiet but consequential divergence in U.S. policy, one that matters deeply for children growing up digital.

On one hand, federal AI policy is accelerating. On the other, federal education policy is fragmenting. Together, these moves reveal a consistent pattern: speed and flexibility for platforms; complexity and burden for public systems that serve children.

🇺🇸 Federal AI Policy: Speed Without Guardrails

On December 11, the White House issued an Executive Order asserting federal supremacy over AI regulation. The stated goal is to prevent states from enforcing laws that might slow innovation, including measures addressing algorithmic bias, transparency, and accountability.

The logic is familiar. A “patchwork” of state rules is framed as an obstacle. Speed is framed as security.

But in education, this logic breaks down.

AI is already embedded in grading, feedback, tutoring, and assessment tools used by millions of students. State-level protections, often the only mechanisms addressing bias, explainability, and student harm, are now legally vulnerable.

The message is clear. Innovation first, accountability later.

🏛️ Federal Education Policy: Fragmentation Without Capacity

At the same time, the U.S. Department of Education finalized interagency agreements transferring administration of its largest K–12 funding streams (Titles I–IV) to other federal agencies, most notably the Department of Labor.

This isn’t decentralization to states. It’s fragmentation within the federal government.

Districts and state agencies must now navigate multiple reporting systems, compliance regimes, and administrative cultures to access funds that were previously coordinated through a single office.

The irony is hard to miss.

The federal government argues that companies cannot reasonably navigate fifty state AI laws, yet it asks schools to navigate four federal agencies to serve children.

Speed for platforms. Complexity for schools.

🌍 Signals Worth Watching

Several developments this week point to broader structural pressures shaping digital childhood:

• AI’s Technical Limits Are Becoming Visible — Leaders at Google DeepMind, OpenAI, and Anthropic are increasingly acknowledging that current AI models are hitting practical limits. Attention is shifting toward continual learning: systems that can learn over time rather than freezing after training. The challenge is severe. New learning often causes models to forget earlier knowledge. A problem known as catastrophic forgetting. This complicates the narrative that scale alone will deliver safe, human-level AI.

• El Salvador & xAI — A national rollout of generative AI tutoring across more than 5,000 public schools highlights the scale of ambition, and unresolved questions about quality, dependency, and critical thinking.

• U.S. vs. China — While the U.S. goes all-in on frontier AI, China is hedging with sustained investment in energy, batteries, and manufacturing. As Tim Wu suggests, infrastructure strategy may matter more than model size.

• The Data Center Boom — Massive private investment in AI infrastructure is diverting labor and resources from public works, delaying roads, bridges, and utilities. Educational futures are tied to physical systems we rarely discuss.

🤔 Consider

If we continue to develop our technology without wisdom or prudence, our servant may prove to be our executioner.

— Omar N. Bradley

This week, digital childhood is being shaped as much by systems we build as by the skills we teach. AI races, data center booms, and fragmented education policy show that technology moves faster than the institutions meant to safeguard children. Innovation without infrastructure or oversight shifts risk onto learners, educators, and communities.

Digital literacy isn’t enough if the systems behind learning (energy, connectivity, governance) cannot support it. Who benefits from speed, and who bears the cost?

⚡ What You Can Do This Week

• Track AI and infrastructure trends: Note local data center expansions, energy pressures, or federal policy shifts. Understand who decides what gets built, and who carries the risk.
• Assess readiness in your community: Examine whether schools, libraries, and youth programs have the tools, training, and capacity to implement AI and digital curricula safely.
• Spot equity gaps: Identify students or groups at risk of being left behind by fragmented systems. Advocate for targeted investments.
• Promote data and digital safety: Encourage robust security and privacy practices in schools and districts. The human cost of breaches is real.
• Connect technology to infrastructure: Ask how educational and policy decisions intersect with physical systems. Networks, servers, energy, and maintenance. Policy alone won’t protect children.

🔗 Navigation

Previous: DL 415 • Next: DL 417 • Archive: 📧 Newsletter

🌱 Connected Concepts

• AI Infrastructure — Private data center booms, continual learning research, and national AI rollouts highlight how computational capacity shapes educational futures.
• Policy vs. Capacity — Federal AI speed versus education fragmentation illustrates the tension between regulation, innovation, and institutional readiness.
• Digital Sovereignty — Who controls the systems and data that children rely on? Federal and international policy choices affect autonomy and risk exposure.
• Equity in Digital Childhood — Unequal infrastructure, access to AI tools, and teacher training deepen disparities in youth digital experiences.
• Contingent Innovation — Technological progress depends on human and physical systems; failures in energy, labor, or governance reveal hidden limits.
• Technological Risk — Omar Bradley’s warning resonates: unchecked innovation without prudence can harm learners and institutions.
• Global AI Strategy — Contrasts between U.S. frontier investment and China’s practical infrastructure focus demonstrate divergent approaches with real implications for youth.
• Educational Resilience — The capacity of schools and districts to absorb technological change, maintain safety, and deliver learning under pressure.