DL 423

The War on Friction

Published: March 1, 2026 • 📧 Newsletter

In 2026, friction has become the villain.

From the Pentagon demanding “frictionless” autonomous systems to lawmakers pushing age verification infrastructure in the name of child safety, guardrails are being recast as inefficiencies. Boundaries are framed as bottlenecks. Deliberation is treated as delay.

But friction is not a design flaw. It is often the only thing keeping power accountable.

When we remove resistance in the name of scale, we don’t eliminate constraint. We relocate it. And usually, we relocate it downward.

This issue isn’t about AI capability. It’s about what happens when we decide that anything slowing acceleration must be stripped away.

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

📚 Recent Work

I posted the following this week:

• The Door Was Never Locked — As i continue to write about digital sovereignty and running your own services, I've had folks asking a simple question. WHY? This post is for you.
• Beyond Zoom: Why We Took Control of Our Video Conferencing with Jitsi Meet — I'm starting to use Jitsi when I need to video conference with others.
• Under the Hood: Setting up a Self-Hosted Jitsi Server on Reclaim Cloud — Here's how I got Jitsi up and running.
• Thinking Out Loud: What Self-Hosting Is Teaching Me About Computational Thinking — How do we reason carefully about risk, tradeoffs, and responsibility when there is no perfectly safe option?
• Chapter 2: SENSE — What Helps You Keep Going? — This week we launched the next chapter of the Signpost Sessions. Moving from naming the weight to noticing the anchor.

🔖 Key Takeaways

• Friction is being rebranded as failure. Guardrails, verification, and limits are increasingly framed as inefficiencies rather than safeguards. Especially when they slow scale or constrain power.
• Accountability doesn’t disappear. It relocates. When companies refuse guardrails, governments impose them. When governments demand removal, companies become leverage points. Power shifts, but the tension remains.
• Extraction is recursive. Foundation models scraped human knowledge. Smaller models distill larger ones. Platforms monetize identity. Everyone claims innovation. Everyone operates downstream of someone else’s data.
• “Protecting children” and “national security” are friction accelerants. They are rhetorically powerful enough to justify structural shifts that would otherwise provoke resistance. Including surveillance infrastructure and market isolation.
• Removing resistance reshapes intelligence. If we eliminate the cognitive and institutional friction that supports deep focus, we shouldn’t be surprised when performance metrics shift.

🛡️ Follow-up: The "Safety" Standstill Shatters

Two weeks ago in DL 421, we discussed whether Anthropic’s "safety-focused" branding was a principle or a positioning statement. This week, we got our answer.

The standoff over a $200 million Pentagon contract ended in a total rupture. Anthropic CEO Dario Amodei refused to remove "guardrails" that prevent Claude from being used for mass domestic surveillance or fully autonomous lethal targeting. In response, the Trump administration didn't just cancel the contract. They went nuclear.

Defense Secretary Pete Hegseth has officially designated Anthropic a "supply chain risk to national security."

Why it matters: This doesn't just end the Pentagon deal. It forces any private company doing business with the military to cut ties with Anthropic. It is a state-sponsored attempt to isolate a major U.S. tech firm from the economy.

🕵️ The Rear-Guard Action: China’s Distillation War

While the U.S. government attacks Anthropic from the front, Chinese firms appear to be probing it from the back.

Anthropic recently disclosed that labs including DeepSeek allegedly used 24,000 fraudulent accounts to conduct what are known as “distillation attacks.” Systematically querying Claude to train smaller competing models on its outputs.

To understand why this matters, we need to be precise about terms.

Distillation is a legitimate AI training technique. A smaller “student” model learns by observing the outputs of a larger “teacher” model. Instead of training on raw internet data (expensive and compute-intensive), the student learns to mimic how the teacher responds. It’s a shortcut.

Inside a company, this is standard optimization. Outside of a company, using mass fake accounts to extract behavioral patterns. It starts to look less like learning and more like copying the exam key.

The Recursive Irony: Foundation AI models (including Claude) were trained on scraped public data created by millions of humans who were never paid.

Everyone in this stack is downstream of extraction. The difference isn’t whether extraction exists. It’s who is allowed to do it, and under what rules.

🆔 KOSMA: The "Safety" Trojan Horse

The pressure on "Big Tech" is moving from the Pentagon to the living room. Senators Ted Cruz and Brian Schatz are reviving the Kids Off Social Media Act (KOSMA).

• The Hook: It bans social media for under-13s and cuts algorithmic feeds for teens.
• The Hidden Cost: Because the bill holds companies liable for "fairly implied" knowledge of a user's age, the only way for platforms to comply is Mass Age Verification.
• The End of Anonymity: To stay legal, the internet is becoming a "Verified Space" where you’ll likely need a government ID or facial scan just to log in. What is sold as "protecting kids" is effectively building the infrastructure for a national digital ID system.

📉 A Cognitive Decline?!?!

Why the sudden rush to “save” the kids? Because we’re finally seeing data from what amounts to a 25-year, $30 billion experiment in classroom technology. It’s once again provoking a predictable cultural panic. Schools nationwide now spend billions on laptops, tablets, and digital infrastructure, yet academic performance has not improved in the ways advocates once promised

In testimony before the U.S. Senate earlier this year, neuroscientist Jared Cooney Horvath argued that Gen Z may be the first generation in modern history to score lower on standardized cognitive performance measures than their parents. This is despite unprecedented access to digital tools and institutional schooling.

A couple of things:

• We've seen this narrative play out again and again as we suggest that "kids are getting dumber."
• That phrasing, “first in modern history” echoes what researchers sometimes call a reverse or negative Flynn Effect, where long-standing increases in measured test scores have plateaued or even reversed in some places.
• Performance on the tests we’ve relied on for decades is slipping, especially in areas tied to deep focus and sustained reasoning.
• What we call “intelligence” is shifting as the educational environment changes.

🔎 Consider

A gem cannot be polished without friction, nor a man perfected without trials.
— Seneca

But we forget that institutions are shaped the same way.

Verification processes. Editorial review. Age limits. Guardrails in AI systems. These are not bureaucratic annoyances. They are polishing mechanisms.

When we remove friction from systems of power, we do not create purity. We create smooth surfaces where nothing can catch. Not oversight, not accountability, not dissent.

The question isn’t whether friction exists. It’s who gets to feel it.

⚡ What You Can Do This Week

• Identify one guardrail you value. Pick one boundary you don’t want removed. Name it explicitly. Write down why it exists. When you can articulate the purpose of a guardrail, you’re less likely to accept its removal as “inevitable.”
• Watch the language shift. This week, we notice when you see certain words (efficiency, optimization, streamlining, modernization, national security, protecting children). Ask what friction is being removed? Who benefits from its removal? Who loses leverage? Language is usually the first place the boundary moves.
• Practice Strategic Slowness. Do one thing this week the slow way. Intentionally. Sit with a claim long enough to feel its implications. Acceleration concentrates power. Slowness redistributes awareness.

🔗 Navigation

Previous: DL 422 • Next: DL 424 • Archive: 📧 Newsletter

🌱 Connected Concepts

• Friction as Infrastructure — the idea that resistance, guardrails, and verification mechanisms are not inefficiencies but structural supports for trust and accountability
• Narrative Capture — how language like “efficiency,” “innovation,” or “protecting children” is used to justify structural shifts in power
• Extraction Stack — the layered system in which data, labor, and behavioral signals are continuously harvested across platforms and AI systems
• Strategic Ambiguity — the gap between what companies or governments publicly claim and what their structural incentives actually produce
• Civic Slowness — intentional resistance to acceleration in order to preserve human agency, institutional deliberation, and democratic oversight
• Boundary Design — how societies decide where limits belong: age limits, content limits, guardrails, and accountability mechanisms