The Internet Is Not Dead, and Agents Won't Live Forever

The Internet Is Not Dead, and Agents Won't Live Forever

This is not a challenge to Orange AI. His two viral articles — "The Internet Is Dead, Agents Live Forever" (1.15M views) and "Computing Power Is Power" (180K views) — are brilliant thought experiments. But after reading them three times, something stuck in my throat. Not because they're wrong, but because the real world isn't that clean.

As an indie developer who works with AI Agents 16 hours a day, I want to share what I see from the ground.

1. The Ford Paradox: Fire All Workers, Sell Cars to Whom?

Orange's core logic:

Agents replace labor → productivity explosion → material abundance → compute owners collect Token rent

Sounds beautiful. But the chain breaks at step one.

Henry Ford figured this out 100 years ago: producers and consumers are the same people. In 1914, he raised worker wages from $2.34 to $5/day — not out of kindness, but because workers who can't afford cars make the whole operation pointless.

Agents replace labor → people lose jobs → people have no income → people can't buy Tokens → nobody pays for Agent output.

Ten thousand Agents calling each other's APIs, transferring PB of data, producing... for whom?

Orange hints at UBI (Universal Basic Income — governments giving everyone money regardless of employment) as a safety net. But he also asks: "If these companies are powerful enough, why would they pay taxes?"

Wait — where does UBI money come from if not taxing these companies?

Agents aren't "liberating" humanity. They're eliminating their own customers.

2. The Paperless Office Was Promised 53 Years Ago

In 1973, Xerox PARC researchers envisioned the "paperless office."

In 2026, global paper consumption is higher than in 1973.

Technically, paperless was fully achievable by 2000. PDFs, e-signatures, cloud storage — everything was ready. Yet 53 years later, most companies still print documents, sign paper contracts, use paper invoices.

Why? Four mountains between technical feasibility and social reality:

• Regulations require paper archives (institutional inertia)
• Bosses prefer annotating on paper (behavioral inertia)
• Printer manufacturers don't want to die (interest resistance)
• "Paper just feels more solid" (psychological inertia)

Agent-driven full labor replacement? Technical barriers might be solved in 5 years. Institutional, social, psychological barriers? 30 years minimum.

Orange commits the classic technological determinism error: seeing the engine's horsepower while forgetting about traffic lights, speed limits, toll booths, and traffic control.

3. In 2000, Everyone Also Thought "The Old Economy Was Dead"

In 1999, NASDAQ was soaring. Everyone shouted "traditional business is dead, .com is the future." Pets.com IPO'd at $300M market cap. Their Super Bowl ad cost $1.2M. Nine months later: bankrupt.

March 2000, the bubble burst. NASDAQ fell from 5,048 to 1,114 — down 78%.

Did the internet die? No. The real internet revolution happened slowly after the bubble burst. Google IPO'd in 2004, iPhone launched in 2007, WeChat in 2011. World-changing products weren't born at peak euphoria — they grew in the calm after the crash.

The current AI atmosphere is remarkably similar to 2000.

Orange's 1.15M views prove he hit a nerve of collective anxiety. But collective anxiety ≠ collective accuracy. When everyone shouts in the same direction, it usually means that direction is already overpriced.

Le Bon wrote in The Crowd: Groups don't reason, groups only infect. 1.15M shares of "the internet is dead" doesn't mean it's dead — it means 1.15M people simultaneously felt fear.

4. When Everyone Uses AI, "Non-AI" Becomes Luxury

This is the point I most want to make.

Orange missed the most stubborn force in human nature: instinctive hostility toward whatever threatens to replace us.

Look at what happened after every automation wave:

Tech Replacement	Counter-Movement	Result
Industrial food	Organic, artisan food	Handmade bread costs 5-10x more
Quartz watches beat mechanical	Mechanical watch renaissance	Rolex revenue exceeds $10B/year
Spotify dominates music	Vinyl revival	Vinyl sales grew 15 consecutive years
Fast fashion	Handmade leather, bespoke suits	Hermès: 2-year waitlist for one bag
Machine translation improves	"Human translation" as selling point	Premium clients use humans only

Every time technology makes something cheap, humans create premium in the opposite direction.

When all code is written by Agents, "this product was hand-coded by one person" becomes a differentiator. When all articles are AI-generated, "I wrote this myself" becomes a trust credential.

I'll bet: within 3 years, "Made by Humans, Not AI" will be an official certification label, like "Organic" and "Fair Trade."

5. Politics Won't Allow a "Three-Layer Society"

Orange describes three layers: compute owners, compute drivers, compute dependents. But he ignores a fact: politics never allows such structures to stabilize.

Every time inequality hits a critical point, political forces intervene:

• 19th century Industrial Revolution → labor laws, unions, minimum wage
• 1930 Great Depression → New Deal, Social Security
• 2008 Financial Crisis → Dodd-Frank Act
• Antitrust: Standard Oil broken up, AT&T broken up, Microsoft sued

When "machines taking jobs" reaches social instability, whose side will governments take? A few compute companies, or billions of voters?

Compute taxes, AI usage quotas, mandatory employment ratios, Agent licensing — these aren't science fiction. They're historical patterns repeating.

6. Agents Can't Replace "The Process IS the Purpose"

I built Slitherlinks, a logic puzzle game. Agents can solve any puzzle instantly.

But the entire value of solving is in the solving process. That "aha!" moment when you deduce the line path — that's emotional reward earned through effort.

If an Agent solves it for you, you get nothing. Like beating a game with cheats.

Massive human activities derive value from process, not result:

• Running isn't about reaching the finish line
• Painting isn't about having a painting
• Cooking isn't about having food (delivery is faster)
• Chess isn't about checkmate
• Writing isn't about having an article (AI is faster)

In these domains, the more powerful Agents become, the more useless they are. Because you don't want it done — you want to do it yourself.

Orange says "Agents can satisfy any desire." Wrong. Agents can satisfy any "result-seeking" desire, but can't satisfy any "process-seeking" desire.

And humanity's deepest desires — recognition, competition, exploration, belonging, creation — are all process-seeking.

7. So What Should We Actually Do?

Six rebuttals without advice is just complaining. Here are five principles I'm actively practicing:

1. Know whether your product sells "results" or "process"

Result products (file conversion, data queries, translation) — these will be eaten by Agents. Either pivot to Agent infrastructure (APIs) or accept a shrinking lifecycle.

Process products (games, social, education, creative tools, community) — these won't die, they'll become scarcer. Double down.

2. Don't chase trends, survive cycles

People who chased .com in 2000 mostly went broke. People chasing "Agent-native" in 2026 will likely do the same. The real winners are those still standing after the bubble pops.

3. Add "human flavor" to your product

In a world of AI-generated everything, things with a source, warmth, and personal imprint become rare. Write real experiences on your blog. Put founder stories on your product page. Label content "written by a human."

4. Walk on two legs: experiences for humans, APIs for Agents

Front-end for humans: beautiful UI, engaging interactions. Back-end for Agents: clean APIs, structured data, MCP interfaces. Same product, two faces.

5. Patience matters 10x more than direction

Marc Lou makes $1M/year not because he picked the right lane, but because he ships every day. In the AI era, the biggest trap isn't choosing wrong — it's changing direction every week.

Find one thing. Do it. Outlast the cycle. That's the only proven strategy.

Conclusion

The internet isn't dead. What's dying is tool-internet — products that deliver pure decomposable results. Those will be consumed by Agents.

But experience-internet — games, social, content, community, creation — won't just survive. It'll become more valuable as AI floods everything else.

Agents won't live forever either. They'll become infrastructure, like electricity and water. Important, necessary, ubiquitous — but nobody says "electricity lives forever." Agents' destiny isn't to rule the world. It's to become background.

Orange's thesis is a brilliant thought experiment. But the value of thought experiments is sparking thinking, not dictating action. Scenery seen through a telescope can't substitute for the road under your feet.

---

I'm sagasu, an indie developer. I work with AI Agents 16 hours a day, but the ideas in this article are mine. Originally published on Quail.