In Defense of Bubbles

Every few weeks, someone leans in with the same hushed certainty: “AI is a bubble.” What they usually mean is more familiar:

• Most pilots fail, so it must be fake.
• Developers slow down on real code, so it must be useless.
• It hallucinates, so it can’t be trusted.
• My team’s workload went up, so it doesn’t work.
• Costs are wild, so nobody’s making money.
• It’s just autocompletion with PR.

Strip away the linkedin/reddit wars and it’s the same resistance we hear during every paradigm shift: roles are changing, methods feel alien, and it’s safer to label the new thing unserious. Yes, it’s messy. That doesn’t make it empty, it makes it early.

We judge bubbles by headlines and forget what they leave behind.

• In the 1840s, Railway Mania wiped out plenty of investors, but Britain kept the tracks, the national logistics grid that powered a century.
• In the dot-com era, froth overbuilt fiber and data centers. The crash forced better methods such as Agile and CI/CD. Cheap bandwidth plus disciplined build gave us cloud, mobile, and the internet we stand on today.

That’s the pattern: frenzy → overshoot → reset → golden age on the leftovers.

AI isn’t just connecting machines like previous technological shifts, it’s rewiring how we work.

Value moves from typing to deciding, from syntax to taste, from “How do I implement?” to “What exactly are we building, and why?”

Enter the vibe coder, akin to a conductor. They set intent, constraints, and feel, and let the machine handle scaffolding.

Chaotic now? Of course. Post froth, the cohort hardens around tests, evals, and governance, just as post dot-com engineers matured into Agile.

Early pilots stumble because exploration is doing its job. Wins show up when you refactor the workflow, not when you duct tape a bot onto a broken pipe.

• Speed with judgment. AI kills boilerplate and hunt time, humans still own architecture, naming, security, and taste.
• Democratization with standards. More people can ship, quality comes from reviews, tests, clean data, and crisp product choices.
• Real work, real gains. Cut the hunt (retrieval), cut the draft (macros or scaffolds), cut the handoffs (smart routing). If CSAT holds and reopens drop, that’s not theater, that’s progress.

Tools don’t build products. Teams with taste do, faster, now.

• After two years of AI headlines, fatigue sets in, so the next wave must be a distraction.
• Quantum timelines are fuzzy, so it must be vapor.
• Today’s products are niche, so critics read “niche now” as “nothing ever.”
• A couple of big raises later, “same VCs, new buzzword.”
• And declaring a pivot lets pundits say, “Told you AI was hype,” without engaging where it’s actually delivering.

Here’s my take: quantum isn’t a bailout for AI, it’s the CAN DO instinct of engineering.

Receipts for CAN DO

• CERN to the Web. A data sharing fix for physicists became HTTP and HTML and rewired the planet.
• Reusable rockets. “You can’t land that.” Then we did, launch costs fell and new orbits of business opened up.
• mRNA. “One day” became “now,” and the platform keeps paying beyond a single vaccine.
• AlphaFold. A moonshot turned into standard lab tooling for materials and medicine.
• Quantum inspired optimization. Even pre fault tolerance, ideas from quantum are already improving routing and scheduling on classical hardware.

• Infrastructure: compute, models, eval harnesses, data pipes, foundation libraries.
• Method: reliable patterns for RAG and agents, guardrails that protect speed, sane governance.
• People: builders who can steer machines with clarity, define problems crisply and ship small, measured, iterative bets.

That’s the soil where durable companies grow.

Bubbles are messy on the surface and compounding underneath. They drag capital, talent, and urgency into hard problems and force us to invent the methods we’ll all use after the dust settles.

So no, I’m not bracing for a collapse. I’m building for the after. The durable value lives in the infrastructure, the method, and the people sharpened while everyone else argues about labels.

Keep the CAN DO. Ship small. Measure hard. Scale what works. When this wave crests, we’ll be standing on the bedrock it left behind.