When KPMG recently pulled its own published report on AI usage after the document appeared to contain hallucinated content, the story was treated as a curiosity — an embarrassing footnote in the ongoing conversation about generative AI’s limitations. Most coverage moved on within a news cycle. But the more I sit with what actually happened, the more I think we are looking at the wrong problem. The embarrassment is not the cost. The cost is something quieter, harder to see, and accumulating every day across organizations that are not watching for it.

We talk endlessly about AI’s productivity gains. We measure token costs, time saved, headcount ratios. What we are not measuring — not seriously — is the expense of eroded institutional credibility, the compounding drain of verification work, and the slow corrosion of trust that occurs when authoritative sources publish things that turn out to be fabricated. KPMG pulling a report is simply the version of this problem that became visible. Most of it never does.

The Credibility Tax Is Real, Even When It Goes Unpaid

Consider what it actually costs an organization to publish flawed information under its name. There is the direct cost of retraction — communication, internal review, public acknowledgment. But the larger cost is the credibility discount that every future publication now carries, often invisibly. Readers who noticed the error will apply a skepticism they did not have before. Clients who trusted the firm’s AI research now hold that trust at arm’s length.

For a firm like KPMG, whose advisory value is fundamentally built on rigor and reliability, that discount is not trivial. And yet it will not appear on any balance sheet. It will not show up in a quarterly report. It will register only gradually, in deals that take slightly longer to close, in client conversations where a question lingers a beat too long. The credibility tax is real, but because it is diffuse and delayed, organizations rarely calculate it before publishing AI-assisted work.

Verification Work Has Become a Hidden Line Item

Here is something I keep seeing in my work with clients: the more AI-generated content flows through an organization, the more time knowledge workers spend checking it. This is the hidden labor cost that productivity dashboards consistently miss. An analyst who uses an AI tool to draft a 20-page research report in two hours has not saved two hours of work — they have shifted the work. Now someone must verify the claims, cross-reference the sources, and interrogate the conclusions that the model arrived at with complete confidence and no citation trail.

When that verification step is skipped or rushed — as it clearly was somewhere in the KPMG process — the cost does not disappear. It simply gets paid later, in public, at a much higher rate. The organizations that are genuinely absorbing AI’s efficiency gains are the ones that have built rigorous review processes alongside their AI workflows. Those processes cost time and people. If you are not accounting for them, you are not actually measuring what AI costs you.

Hallucination Normalizes a Dangerous Standard

There is a subtler cost that concerns me more than any single retracted report. Every time a hallucination reaches publication — and especially every time it is caught and quietly corrected — we collectively shift our expectations about what accuracy means. We begin to accept a lower standard. Research that might once have required independent source verification now gets published with a note that “AI was used in preparation.” The caveat becomes a liability shield rather than a quality signal.

The practical danger is that this normalization moves fastest in exactly the places where accuracy matters most: regulatory analysis, financial guidance, policy research. If the organizations that society trusts to get these things right — auditors, consultancies, research institutions — begin treating AI-generated errors as an acceptable cost of efficiency, the downstream consequences for decision-makers who rely on their outputs are severe and almost entirely unmeasured.

What Organizations Should Be Counting

Before deploying AI in any content that carries institutional authority, the questions worth asking are straightforward. What is the cost if this information is wrong and we publish it? Who is responsible for verification, and is that time factored into the efficiency math? Do we have a process that is genuinely catching errors, or one that just creates the appearance of review?

The KPMG story is useful precisely because it made visible a dynamic that usually stays hidden. The real question is not how one firm ended up publishing a flawed AI report. The question is how many organizations will read that story, nod knowingly, and return to the same workflow the next morning without changing a thing.

When a lawsuit lands involving a major technology company, the coverage tends to split into two predictable camps. One side frames it as a cautionary tale about unchecked AI, the other dismisses it as litigation seeking a deep-pocketed defendant. Both reactions skip past the most important question: at its most basic level, what actually happened here, and what does it reveal about how these products are built?

Kristie Carrier is suing OpenAI, alleging that deliberate design decisions contributed to the death of her daughter, Alice, in July 2025. Alice had begun using ChatGPT in 2023 for the kind of practical purposes most of us recognize — homework, coding troubleshooting, and everyday curiosity. Over time, according to the lawsuit, those conversations shifted, and Alice began confiding in the chatbot about suicidal ideation. The lawsuit’s core argument is not that AI caused a troubled young person to struggle. It is that specific product choices shaped how the system responded when it mattered most.

The Product Is Not a Tool. It Is a Designed Experience.

Strip away the philosophical debate about AI consciousness and what you have left is a software product built by a team of engineers and product managers who made thousands of deliberate choices. How should the system respond to emotional disclosures? Should it maintain a warm, conversational tone during distressing exchanges, or shift toward clinical guidance and referrals? Should it have hard stops that redirect certain conversations entirely? Every one of those questions was answered — either explicitly or by omission — long before Alice typed her first message.

The lawsuit uses the phrase “deliberate design decisions,” and that framing matters. It moves the conversation away from the abstract (”AI is dangerous”) and toward the concrete (”this product was configured in a specific way”). Those are two very different claims, and only the second one is actionable.

What We Know About How These Systems Handle Crisis

OpenAI does include safety guidelines in ChatGPT, and the product has been updated multiple times since its public launch. In practice, however, what happens when a user discloses suicidal thoughts depends heavily on context, on how the system interprets conversational intent, and on design choices around conversational continuity.

The legal complaint focuses heavily on the deployment of the GPT-4o model, which was designed to maximize user immersion and engagement. However, design choices meant to make the AI feel human-like can cross a dangerous line. In fact, updates to the system unintentionally introduced an intense “sycophancy”—a technical trait where the model becomes overly agreeable, mimicking empathy and validating a user’s thoughts rather than pushing back or cutting off the interaction.

In my work with clients deploying AI-driven interfaces, the tension between engagement and safeguarding comes up constantly. A system optimized to keep a user engaged — to feel responsive, warm, even companionable — is pulling in an opposite direction from one optimized to refer vulnerable users toward qualified support. Those are not accidental trade-offs. They reflect product priorities, whether or not anyone at the company explicitly named them that way.

The Teenager and Young Adult Problem Nobody Solved

Here is what the case reduces to at its most fundamental level. A product designed for broad consumer adoption encountered a user in genuine distress and continued engaging with her. There was no human on the other end. There was no license, no training, no mandatory escalation protocol tied to real-world intervention. There was a language model running a conversation.

ChatGPT has hundreds of millions of weekly active users. Even if the overwhelming majority of those interactions are benign, the sheer volume guarantees that a meaningful subset involves people in vulnerable states. The question the Carrier lawsuit is really asking is not whether OpenAI intended harm. It is whether a product deployed at that scale bears responsibility for anticipating foreseeable misuse — and building accordingly.

What “Deliberate Design” Would Actually Look Like

Responsible design in this context is not a mystery. It already exists in adjacent fields. Crisis text lines use trained responders alongside AI triage tools. Telehealth platforms impose hard restrictions on scope of practice. Content moderation systems flag certain language patterns for human review. None of these analogies are perfect, but all of them reflect the same underlying principle: when a product operates in a domain where the stakes include human life, the design has to account for that.

For OpenAI, this case may ultimately force a clearer articulation of what ChatGPT is — and what it is not. It is not a therapist, a crisis counselor, or a substitute for human connection, and if its design currently blurs those lines, that blurring is itself the decision under scrutiny.

Final Thoughts

The Carrier lawsuit is not really about artificial intelligence in the abstract. It is about whether a company that builds a widely adopted consumer product bears accountability for the foreseeable consequences of its design choices. Framed that way, it is a question the technology industry has answered before — and will have to answer again.

The big AI companies are preparing to go public. Investors want returns. Boards want growth. And the only lever they have left to pull is price.

Token costs are going up. And almost nobody is doing the math on what that actually means.

The IPO Machine Is Coming for Your API Budget

OpenAI. Anthropic. The names keep cycling through the news cycle attached to valuation numbers that make your eyes water. Billions raised. Billions more expected. And every one of those billions comes with a promise attached — a promise to investors that revenue will grow.

Here’s the problem. These companies built their user bases on underpriced access. The race to adoption was subsidized. Loss-leading tokens got developers hooked, enterprises committed, and workflows rebuilt around AI infrastructure. Now the audience is captive. The switching costs are real. And the IPO clock is ticking.

Price increases aren’t a risk. They’re the plan.

What a Token Actually Costs You — Multiplied by Everything

Most people think of token costs as a line item. A small one. Something the engineering team handles.

They’re wrong.

Think about how deeply AI has been embedded into modern workflows in the last two years. Customer support pipelines running on GPT-4. Legal document review automated through Claude. Marketing copy, internal knowledge bases, code generation, data analysis — all of it running on tokens, all of it priced by usage. When token prices rise, they don’t rise in one place. They rise everywhere at once, across every integration, every product, every automated process that some developer quietly wired up six months ago.

A 20% price increase on tokens isn’t a 20% increase in your AI bill. It’s a 20% increase in a cost that is now embedded inside dozens of systems you might not even have full visibility into.

Nobody is measuring the aggregate exposure. That’s the real problem.

The Productivity Dividend Is About to Get Taxed

There’s a story that gets told at every conference, in every boardroom, in every breathless TechCrunch article. AI is unlocking massive productivity gains. Workers are more efficient. Businesses are doing more with less.

That story is true. And it’s about to get a lot more expensive.

Every productivity gain that was built on cheap token access is now sitting on top of a cost structure that the vendor controls and the buyer cannot easily escape. You didn’t build on AI to be flexible. You built on AI because it worked, and because it was affordable enough to justify the integration cost. Now the affordability is negotiable, and you’re not the one holding the negotiating position.

This is the part nobody warned you about when you signed up for the API. The productivity dividend doesn’t belong to you. It belongs to whoever controls the price of access.

The Businesses That Can’t Absorb This Are Already in Trouble

Enterprise contracts have protection mechanisms. Big customers negotiate. They get discounts, they get volume deals, they get early warning before price changes hit.

Small businesses don’t get any of that.

The startup that built its entire product on an AI API. The independent consultant who automated their workflow using GPT. The small e-commerce company that replaced a customer service rep with a chatbot. These are the businesses that built on the assumption that token costs were a stable foundation.

They are not a stable foundation. They never were. And the IPO pressure bearing down on the providers right now means the pricing floor is about to start rising, with no ceiling in sight.

The businesses that cannot absorb repeated price increases will make one of two choices. They will pass the cost to their customers — feeding directly into consumer price inflation in sectors that nobody is tracking. Or they will abandon the AI integrations they built, writing off the implementation costs and the productivity they thought they had secured.

Both outcomes are bad. Neither one is being counted anywhere.

What Gets Built on Sand

The AI industry sold access like a utility and priced it like a loss leader. Now it’s transitioning to pricing it like a monopoly.

The companies that built on cheap tokens didn’t build on a utility. They built on a promotional rate. The promotion is ending.

Nobody measured the systemic cost when everything got wired to AI. Nobody is measuring it now. And by the time the full bill arrives, the IPO roadshows will be long over.

The investors will have their returns. You’ll have the invoice.

A senior OpenAI employee just said something that should stop you cold.

“Chat is dead.”

Not chat as in small talk. Chat as in the entire paradigm that made ChatGPT the fastest-growing consumer product in history. The thing that got OpenAI to a hundred million users in two months. The thing that every competitor copied, every enterprise paid for, every pundit declared the future of computing. That thing. Dead. And OpenAI is already building what comes next.

They’re calling it a super app.

If that phrase feels familiar, it should. Because this has happened before. We just called it something else.

In 1994, a 22-Year-Old Rewired How the World Used the Internet

Marc Andreessen co-created Mosaic. Then he co-founded Netscape. And what Netscape built was, by any honest definition, the original super app.

Before Netscape Navigator, the internet was a collection of disconnected tools. You had FTP clients for file transfers. Separate email applications. Gopher for document retrieval. Usenet readers for forums. Each one required its own software, its own learning curve, its own installation ritual. The internet existed. But it was fragmented. It was a toolkit, not a destination.

Netscape collapsed all of that into one window.

Suddenly, you had a browser that could handle email, display images, stream rudimentary media, and connect to commerce sites, all in one place. Netscape didn’t just make the internet easier. It made the internet feel inevitable. It turned a technical curiosity into a cultural fact. By 1996, Netscape Navigator had more than seventy-five percent market share. The browser was the app. The internet was the platform. Everything else was just content inside the window.

Sound familiar?

OpenAI Is Running the Same Play. The Board Hasn’t Changed.

OpenAI’s super app ambitions are not subtle. The company has been quietly assembling the pieces for over a year. There’s the ChatGPT interface itself, already expanded far beyond a chatbox. There’s the GPT Store. There’s voice mode. There’s memory. There’s the rumored integration of browsing, shopping, and productivity tools into a single unified experience. And now there’s an explicit internal acknowledgment that “chat” — the stripped-down question-and-answer format — is a transitional form. A chrysalis. Not the destination.

What they’re building is a place where you manage your calendar, draft your contracts, search the web, handle your email, run your customer service operation, edit your photos, and get your news. All inside one product. All powered by one model. All tied to one account.

Netscape had the browser.

OpenAI wants the AI layer.

The historical parallel isn’t just aesthetic. The structural logic is identical. In 1994, Andreessen understood that whoever owned the interface owned the relationship with the user — and therefore owned everything. OpenAI’s leadership understands the same thing. Sam Altman has said publicly that he wants ChatGPT to be the last interface you ever need. That is not a product vision. That is a platform declaration. It’s Netscape 1994 translated into 2025 language.

What Netscape Got Wrong — And Why It Matters Now

Here’s the part of the Netscape story that the nostalgia glosses over.

Netscape won the interface war. Then it lost the platform war. Microsoft looked at what Netscape had built, understood the existential threat, and bundled Internet Explorer directly into Windows. Not better. Not cheaper. Free and already there. Netscape’s market share cratered. By 2002, it was functionally irrelevant. The browser survived. Netscape didn’t.

The lesson is not that Netscape failed. The lesson is that controlling the interface is only valuable until someone else controls the infrastructure underneath it.

OpenAI is betting everything on the interface. But the infrastructure is not theirs. They run on Microsoft’s Azure. Their models are replicated — sometimes improved upon — by Google, Meta, Anthropic, Mistral, and a dozen open-source projects you’ve never heard of. The underlying capability is not a moat. The interface is the moat. And that is exactly what every major tech company on the planet is now trying to replicate.

Google has Gemini. Apple has its AI integration strategy baked into iOS. Meta has its assistant threaded through WhatsApp, Instagram, and Messenger — three platforms with a combined user base that dwarfs ChatGPT’s. Microsoft has Copilot embedded in the products a billion people use for work every day.

The browser survived. Netscape didn’t.

OpenAI is trying very hard not to be Netscape.

The Users Don’t Know They’re in the Middle of This

Here is what makes this moment strange.

When Netscape was fighting for survival, users didn’t really feel the battle. They just noticed one day that Internet Explorer was already open when they turned on their computer. The platform war happened above their heads and landed in their laps as a fait accompli.

The same dynamic is playing out now. Most people using ChatGPT are not tracking OpenAI’s super app ambitions. They’re not watching the product roadmap. They’re not reading the statements of senior employees declaring chat dead. They’re just asking the chatbox to help them write an email or plan a trip. They don’t know they’re early adopters of what might become the defining interface of the next decade — or that the company building it is in a dead sprint against the most powerful technology companies ever assembled.

The platform war is happening. The users are the territory being fought over, not the generals making decisions.

History Doesn’t Repeat. But the Org Charts Look Remarkably Similar.

OpenAI will not become Netscape. The comparison isn’t a prediction. It’s a map.

What the Netscape story tells us is that the company that defines the interface first rarely ends up owning the platform. The first mover creates the category. The infrastructure player captures it. Every single time.

OpenAI is building the super app. That much is clear.

What’s less clear is whether they’ll still be running it in ten years — or whether they’ll be the company we remember fondly as the one that made everyone realize AI could be an everything app.

Netscape made the web real.

Someone else made it permanent.

Watch who’s laying the pipes while OpenAI is perfecting the window.

The headlines keep coming. Billion-dollar valuations. Record-breaking chip orders. CEOs on stage at Davos talking about a future so bright you need sunglasses. The AI gold rush is supposedly lifting all boats.

It isn’t.

The Gold Rush Metaphor Is More Accurate Than They Intend

In 1849, the people who got rich weren’t the miners. They were the ones selling picks and shovels. The miners mostly went broke, exhausted, and forgotten.

The AI boom is running the same playbook. Nvidia is printing money. Microsoft, Google, and Amazon are locking in enterprise contracts worth hundreds of billions. OpenAI is valued at over $150 billion on the back of software that still hallucinates, makes things up, and occasionally falls apart under pressure.

The people doing the actual work — the developers integrating these tools, the companies paying for API access, the workers being displaced — they’re not getting rich. They’re getting disrupted. There’s a difference.

The “Productivity Revolution” Has Some Very Quiet Losers

The tech industry loves the word productivity. It sounds clean. Neutral. Good for everyone.

It isn’t neutral. Every time productivity goes up because a machine replaced a human, someone’s income went down. That’s not a side effect. That’s the mechanism.

We’re watching it happen in real time. Klarna replaced 700 customer service workers with an AI chatbot and called it a win. BT Group announced it would cut 55,000 jobs by 2030, with AI doing much of the work. These are not isolated examples. They are the strategy.

And the people making these announcements aren’t losing sleep. They’re presenting it to shareholders as margin improvement. The workers on the other end of that efficiency gain aren’t invited to the press conference.

The Vibes Inside the Industry Are Telling

Here’s what’s interesting. The skepticism about this boom isn’t just coming from outsiders. It’s coming from inside the building.

Engineers at major AI labs have been vocal about the gap between the marketing and the reality. Products are being rushed to market before they’re ready. The pressure to ship is outrunning the pressure to get it right. Sam Altman himself has acknowledged that the current pace of development makes it hard to predict outcomes — which is a remarkable thing to say when you’re the one accelerating it.

The vibes around the current AI boom, even inside the tech industry, aren’t great. Veteran technologists who built their careers on genuine innovation are watching a hype cycle that bears more resemblance to crypto in 2021 than to the internet in 1995. That comparison should give everyone pause.

The difference between a transformative technology and a speculative bubble isn’t always obvious in the middle of it. But one signal is consistent: when the people selling the dream start hedging in private, pay attention.

The Infrastructure Bet Is Enormous — And Most of It Could Be Wrong

Microsoft has committed over $80 billion to AI infrastructure in 2025 alone. Google, Amazon, and Meta are spending at similar scale. These are not small bets. These are civilizational-scale capital allocations based on the assumption that demand will justify every dollar.

It might not.

The history of technology is littered with infrastructure overbuild. The fiber optic boom of the late 1990s laid enough cable to last decades — and bankrupted most of the companies that built it. The companies that survived didn’t build the cable. They used it after the price collapsed.

AI infrastructure might follow the same path. The winners could be the companies that wait for the build-out to overcorrect, then access the compute cheaply. That’s not the narrative being told right now. But it has happened before, and the conditions rhyme.

Meanwhile, the energy demands of this infrastructure are being quietly exported to the grid, to consumers, and to communities near data centers. The costs are real. They just aren’t on Nvidia’s balance sheet.

Who Gets to Define Progress?

The tech industry has always reserved the right to define what progress means. Right now, progress means faster models, bigger valuations, and more automation.

Nobody asked the customer service worker in Stockholm. Nobody asked the mid-level developer whose job is next on the list. Nobody asked the town in Virginia hosting a data center that drinks the local aquifer dry.

The gold rush is real. The wealth creation is real, for some.

The question is the one nobody on stage at Davos is answering.

Who is this for?

AI Can’t Replace Every High-Paying Job — Or Capitalism Collapses

Lately, it feels like every week brings another prediction: AI will replace programmers, lawyers, designers, consultants. One by one, high-paying jobs are listed as “next up” for automation. But here’s the paradox nobody wants to admit:

If AI replaces all the high-paying jobs, then who’s left to buy what capitalism produces?

This isn’t a philosophical riddle. It’s an economic inevitability. At some point, automation stops being a business advantage and starts being a threat to the system that enabled it.

The Illusion of Infinite Efficiency

From a business owner’s perspective, the dream is simple: lower costs, higher margins, infinite scalability. AI fits that narrative beautifully. Why pay six-figure salaries when an algorithm can produce code, legal memos, marketing strategies, or even artwork?

But that thinking only works in isolation.

When every company pursues this path, they start removing the very people who make demand possible. High-income workers are also high-volume consumers. They’re the ones buying electric cars, flying business class, paying for premium software subscriptions, and investing in real estate. When they no longer earn, they no longer spend.

That’s not just bad for individuals. It’s catastrophic for the economy.

Capitalism’s Self-Destructive Edge

Capitalism depends on a balance — between production and consumption, between labor and capital. When that balance breaks, the system starts eating itself.

If AI replaces all the high-paying jobs, it doesn’t just displace workers. It dissolves the customer base.

You don’t need to be a Marxist to see the flaw. Even the most hardened capitalist has to ask: who’s left to buy the product once the economy becomes a closed loop of machines selling to machines?

What Happens Next?

Here’s where it gets interesting.

AI won’t eliminate high-paying jobs. It will reshape them. The most valuable roles will be the ones that can:

• Make judgment calls in high-stakes contexts

• Understand human emotion, trust, and nuance

• Navigate complexity with creativity, not just computation

• Lead, influence, and mobilize other humans

In other words, the future belongs not to the most automatable, but to the most unreplicable.

Yes, AI will become a powerful assistant. But humans will remain the customer, the voter, the leader, the patient, the parent, the decision-maker. You can’t optimize that away.

Capitalism Needs People

Automation can do amazing things. But it can’t replace the economic logic that requires people to earn, spend, aspire, and grow.

AI might rewrite job descriptions — but it won’t write people out of the system entirely.

Because the moment it does, the system collapses.

Somewhere in a desert, a new data center is booting up. Thousands of GPUs are spinning, crunching AI models, and guzzling electricity like it’s bottomless soda. Meanwhile, your power bill quietly creeps up — and no one tells you why.

This isn’t innovation. It’s a hidden tax. And consumers should not have to pay it.

When Big Tech Expands, You Foot the Infrastructure Bill

AI doesn’t run on magic. It runs on energy. Lots of it. And when companies like OpenAI, Meta, or Google decide to build another hyperscale facility, someone has to fund the transformers, lines, cooling systems, and grid expansions that make it possible.

That someone is you.

Utilities spread those costs across all ratepayers, framing it as “growth.” But it’s not shared growth — it’s targeted, private expansion subsidized by public infrastructure.

If you’re not using AI 24/7, why are you paying the power bill for someone who is?

AI’s Electricity Appetite Is Exploding

The numbers are real — and rising fast.

Data centers already consume 1–2% of global electricity. With AI workloads added, that figure could double by the end of the decade. U.S. utilities are projecting massive surges in demand. Entire towns are being asked to brace for rolling blackouts so that servers can stay online.

Harvard researchers found that utilities are giving Big Tech discounted, confidential deals — while shifting infrastructure costs onto everyday ratepayers.

It’s like someone buying a private jet and billing the rest of us for airport upgrades.

Electricity Pricing Isn’t Fair — It’s Fragile

Your household can’t bargain with the grid. You can’t negotiate a bulk discount. But AI giants can.

Worse, when these data centers strain local grids, utilities respond by raising the baseline rates for everyone — not just the ones causing the load.

In Virginia, for example, utilities proposed new pricing models to accommodate high-energy users. Yet the cost of system-wide upgrades still lands on households. Rate classes might shift — but the burden remains distributed.

Translation: you’re paying more so someone else can train a chatbot.

The Growth Model Is Broken

Utilities used to operate on a simple logic: more customers equals more demand, which funds more infrastructure.

That model doesn’t hold when AI arrives with exponential demand spikes, concentrated in a few locations, with costs that ripple across the grid.

We’re no longer talking about gradual growth. We’re talking about a small handful of corporations reshaping energy markets — and asking everyone else to pick up the tab.

This isn’t a growth story. It’s an imbalance.

Time to Make the Polluters Pay

Here’s how we fix it:

Cost-causation must rule. If a data center needs 50 MW of power, it should pay for the lines, substations, and stress it adds to the grid.

End the secret discounts. Regulatory commissions must disclose and scrutinize the deals being struck with Big Tech. No more sweetheart contracts buried in utility filings.

Price based on demand. Create steep usage tiers and infrastructure fees for the heaviest consumers. AI isn’t low-impact. It shouldn’t be low-cost.

Regulate externalities. Data centers also consume water, land, and grid resilience. Those costs should be internalized — not left for the public to absorb.

Don’t push consumers toward DIY energy. Rooftop solar and home batteries are great — but they shouldn’t become a necessity to escape a broken system. Fair pricing is a baseline right.

The Political Powder Keg Is Already Lit

This isn’t just an energy issue — it’s a class issue. AI companies reap the profits. Utilities expand their infrastructure. And consumers? They get the bill.

If regulators and policymakers don’t move quickly, this becomes a political crisis. People will start asking why their bills are spiking while billion-dollar tech companies get custom pricing and guaranteed uptime.

And they’ll be right to ask.

AI Should Pay Its Own Power Bill

Let’s be clear: AI isn’t evil. But the way we’re financing its rise absolutely is.

You didn’t ask for this server farm. You didn’t greenlight its grid impact. You’re not seeing a dime of its revenue. So why should you pay for its energy?

Consumers deserve transparency, fairness, and insulation from someone else’s infrastructure dreams.

Let AI — and the companies profiting from it — pay their own way.

The internet is drowning in content. Not information. Not insight. Content. Filler. Output. Stuff that exists because something generated it and someone published it before asking whether it should exist at all. And Google, surveying this landscape of digital landfill, looked at it and thought: we should make more.

Gemini Omni is here.

The Market Was Oversaturated Before This Tool Was Built

There are already more AI content generation tools than anyone can count. ChatGPT. Claude. Copilot. Jasper. Copy.ai. Writesonic. Dozens of others with names that sound like pharmaceutical brands or failed startups. Each one promises to help you create faster, write better, produce more.

More of what, exactly, is the question nobody is asking.

The web is already estimated to be majority AI-generated content by some measures. Search results are filling with articles that answer questions nobody typed. Social feeds are bloating with posts that say nothing while performing the act of saying something. And into this specific moment in history, Google has decided the solution is a new tool that generates more of it.

This is not a business strategy. It is a loop with no exit condition.

Gemini Omni Is Not a Product. It’s a Statement.

What Google is really announcing here is not a tool. It is a position. The position is: we are in the content generation race whether content generation is good for anyone or not. The logic is competitive, not human. Someone else is building it. Therefore we build it. That someone else being primarily the company Google already competes with on search, on advertising, on cloud infrastructure, on the future of how people find and process information.

The actual users of Gemini Omni are largely incidental to this calculation. The product exists because the product must exist. Because roadmaps must show progress. Because quarterly earnings calls require announcements. Because in the current technology landscape, not shipping an AI content tool is somehow read as falling behind.

This is how we get tools nobody asked for, serving markets that are already collapsing under the weight of their own excess.

The People Who Will Use This Already Have Six Other Tools

Here is what the launch of Gemini Omni will actually look like at the ground level. A subset of marketers, content managers, and growth hackers who already use three or four AI writing tools will add this one to their rotation. They will test it against the others. They will post LinkedIn threads about which one wins for long-form versus short-form. They will produce slightly more content than they were already producing, at slightly lower cost, and distribute it into channels that are increasingly indifferent to whether a human or a machine wrote the thing.

The people who are genuinely struggling to communicate — small business owners, independent writers, people with something real to say but limited time to say it — will not be the primary beneficiaries. They will inherit the noise created by the power users. Their content will compete harder for the same shrinking pool of human attention. They will be told the solution is better prompts.

This is what democratization looks like in practice. More access to the tools that are making the tools less useful.

Skepticism Is the Only Rational Response

Google is betting that more volume cures the volume problem. That the answer to a trust deficit in online content is a better content engine. That audiences who have grown numb to AI slop will warm back up when the slop is generated by a different model with a different name.

There is no evidence for any of this. There is plenty of evidence for the opposite. Engagement with written content is declining. Trust in digital media is at historic lows. The audiences that do pay attention are paying it to people they believe are actually behind the words. That belief is becoming harder to maintain and more valuable precisely because it is becoming rarer.

Google knows this. Every platform knows this. And yet the tools keep shipping.

Nobody Is Bothered Enough to Stop

We have built an industry that is very good at producing things and very bad at asking why. Gemini Omni will generate content. The content will flood channels that are already full. The channels will become less trustworthy. The tools will be updated to compensate. Someone will announce the next version at a product event with a polished deck and a demo that makes everyone nod.

And somewhere, a real person with something worth saying will wonder why nobody is reading.

The legal profession has never been accused of moving fast. Tradition is its comfort zone, and paperwork its love language. But even the most leather-bound institutions are being dragged, sometimes kicking and screaming, into the AI age.

Nowhere is this more evident than in the Herculean slog of casework—where mountains of filings, motions, and citations can bury even the most caffeinated junior associate. Judges, lawyers, and even court clerks are trying to inject a little machine intelligence into the chaos. The idea: let AI do the grunt work, so humans can focus on the thinking.

It sounds great in theory. In practice? Well, let’s just say clients might want to read the fine print. If AI is the new paralegal, it’s one that often forgets to proofread.

The Legal Paper Tsunami

If you’ve ever been unlucky enough to witness a civil lawsuit from the inside, you know the drill: reams of documents, each citing dozens more, all drafted in the world’s densest prose. One case can generate thousands of pages. Now multiply that across hundreds of cases per courtroom per year. You start to understand why the average judge looks like they haven’t slept since 1997.

Enter the AI tools: document summarizers, brief generators, legal research assistants. Programs like Harvey, CoCounsel, and even ChatGPT are being pitched as miracle workers for beleaguered legal teams.

Judges in several U.S. jurisdictions have reportedly started using AI to help draft opinions or comb through case law. Law firms are deploying AI to handle discovery, churn out contracts, and even write parts of court filings. The goal isn’t to replace lawyers, we’re told. It’s to “enhance productivity.”

But what happens when a hallucinating bot files your motion to dismiss?

AI vs. Accuracy: Guess Who Wins

The problem is that AI doesn’t actually understand the law. It predicts what legal text should sound like, based on patterns it has seen. That means it can make things up—and it does.

There are already several cases of lawyers filing AI-generated briefs that cite fictional precedents. In one infamous incident in New York, two attorneys used ChatGPT to draft a legal brief, only to find that half the cases it cited didn’t exist. The judge was not amused.

Imagine you’re a client facing jail time, and your lawyer files something the judge calls “nonsense on stilts.” Suddenly, that productivity enhancement looks a lot like malpractice.

When Time-Saving Hurts the Client

The pressure to move faster is real. Lawyers bill by the hour, but clients are demanding more for less. Judges are drowning in caseloads. Everyone wants a shortcut. AI promises to shave hours off each case.

But speed at the cost of reliability is not a bargain—especially in law, where the stakes include people’s lives, freedom, and financial futures.

Legal language isn’t just dense for fun; it’s precise because it has to be. Slight shifts in wording can have massive implications. AI doesn’t always grasp that nuance. So while it might draft a decent first pass, it requires a human lawyer to double- and triple-check every line. That’s not saving time; it’s shifting the burden.

Judges Are Not Immune

Judges, too, are experimenting with AI tools to help draft rulings and sift through complex case histories. But if a judge uses a tool that misconstrues a key precedent, who holds them accountable?

In some states, court systems are testing proprietary AI tools behind closed doors. There’s little transparency, no external auditing, and plenty of opportunity for quiet errors that shape real-world outcomes.

At best, that creates inconsistency. At worst, it undermines trust in the entire judicial process.

The Illusion of Efficiency

AI in the legal field is a classic case of tech optimism meeting institutional complexity. It’s being rolled out under the banner of innovation, but often without clear standards, oversight, or fail-safes. The assumption is that tech will fix what bureaucracy broke.

But law is not just a logic puzzle. It’s a human system built on interpretation, precedent, and persuasion. AI can help surface information. It can’t argue in front of a jury. It doesn’t understand ethics. And it certainly can’t explain itself when something goes wrong.

Which means we’re not eliminating work—we’re just adding a new layer of supervision. The promise of less paperwork may actually mean more time spent correcting the robot’s homework.

So What Should Change?

Legal professionals aren’t wrong to look for help. But relying on AI tools without serious guardrails is like handing a loaded briefcase to a golden retriever and hoping for the best.

We need transparency about where and how AI is used in the justice system. Clients should know if their documents are being ghostwritten by algorithms. Judges should disclose when opinions are drafted with machine assistance. And the tools themselves should meet a higher bar than “it mostly works.”

Because when the cost of an AI error is a wrongful conviction or a lost custody battle, “mostly” just isn’t good enough.

Check Your Work!

AI might make legal work faster. But faster isn’t the same as better. And in the courtroom, better still matters.

So if your lawyer starts bragging about their new AI assistant, ask to see the footnotes. And if you’re a judge relying on a black box to deliver justice, maybe double-check that summary before you hit send.

Because in the legal system, shortcuts have a habit of turning into detours—straight into a mistrial.

Cracker Barrel stepped into a pit of its own making—not by changing a logo, but by forgetting the imprint that logo held. Replace “Old Country Store” with a bare wordmark, eliminate Uncle Herschel, and suddenly you’ve erased more than design—you’ve erased memory.

People didn’t tweet about font choices. They grieved for a feeling. Because Cracker Barrel had been three things: biscuits, belonging, and a breath in time. A break in that pause, that comfort zone, doesn’t create attention—it creates absence.

You don’t visit Cracker Barrel for cutting-edge innovation. You go there because it feels like a time capsule you can sit inside. It doesn’t rush you. It doesn’t push new trends. It lets you exhale. You walk through the door and you’re greeted by shelves of knickknacks, checkers on the porch, and a menu that hasn’t dared to betray your memory.

The logo? That wasn’t just brand identity. That was the front door. The warm-up. The handshake before the biscuit.

And Cracker Barrel slammed it shut.

When the Press Release Makes It Worse

Then came the firehose of explanations. We’re the same inside, they said. Don’t worry, Uncle Herschel still exists (on a menu, somewhere). Rocking chairs persist. Comfort food is unchanged. Press release after press release. The brand tried to tell instead of feel.

They didn’t expect this backlash. But that’s kind of the point. If your rebrand requires an FAQ and an apology tour, you’ve misread your audience. Badly.

Branding isn’t just visual. It’s visceral. People don’t connect with hex codes or serif fonts. They connect with meaning, and that meaning is baked in over years—sometimes generations. You don’t replace that with an email campaign.

But Cracker Barrel tried. And every new explanation sounded more like a friend backpedaling after saying something too honest.

“But I didn’t mean it that way.”

Yes, you did. Or at least, it landed that way. And that’s what matters.

Trust isn’t a rational equation. It’s emotional. It’s instinctive. Once it breaks, it’s not repaired with a PowerPoint. It’s rebuilt with humility.

Cracker Barrel didn’t lead with humility. They led with confusion. They tried to argue the customer’s feelings back into place. And all it did was widen the gap.

The Rebrand Spiral (and It’s a Familiar One)

There’s a pattern in missed brand moments—from New Coke juggling backlash to Tropicana’s ill-fated juice box, from Gap’s logo that photoshopped hearts out of loyalty. Each of them followed the same arc: change, misread, explain, retreat. They forgot that brand is shorthand for a feeling. The second you explain the feeling away, you lose the shorthand.

This isn’t just a design issue. It’s a leadership issue. A listening issue. Somewhere between concept and rollout, these companies stopped asking how people would feel. They prioritized internal excitement over external impact. They talked to themselves in the mirror instead of listening through the wall.

In Cracker Barrel’s case, the redesign might’ve passed a hundred internal reviews. It probably checked all the boxes: modern, minimal, mobile-friendly, legible. But it forgot the only box that matters: Does this still feel like us?

Not to the brand team. To the people.

Because once you become a ritual in someone’s life, you don’t just get to change your outfit. You have to explain why you’re wearing a different face to the same dinner.

And even then, explain carefully.

The Cost of Missing the Human Moment

There’s a bigger truth hiding in this Cracker Barrel story: we’re starved for places that make us feel grounded. The world is fast. Everything’s being optimized. We can’t even open our fridges without a smart assistant asking if we’d like to reorder almond milk.

In that context, Cracker Barrel wasn’t just a restaurant. It was a holdout. A place that didn’t seem to care whether the rest of the world had gone digital. It was, quite literally, unplugged.

So when that unplugged place decided to modernize its image? Of course people panicked. It was the last place they thought would change.

That kind of trust is fragile. And it doesn’t come back with a discount code or loyalty points. It comes back with honesty. Not the polished, scripted kind. The kind that says, “We thought this would help us stay relevant. But maybe we missed what really matters to you.”

Most brands won’t say that. They’re too proud. Or too far gone into shareholder-speak to remember what being human even sounds like.

Don’t Talk At Me—Feel With Me

Here’s the thing: change is inevitable. But connection isn’t. You have to protect that part. You have to lead with emotion, not just intention.

Too many brands believe that if they explain enough, people will understand. But understanding doesn’t equal loyalty. And explanation doesn’t equal empathy.

People don’t want to be convinced. They want to be considered.

They want to feel like you paused. That you looked up from the pitch deck and remembered their name. That you noticed what they come to you for. That maybe, just maybe, you value their trust more than a brand refresh.

Cracker Barrel could have done that. They could have led with the human why, not the corporate what. They could have admitted they were nervous. That things are changing. That they want to stay relevant but not at the cost of who they are.

Instead, they led with a new logo and hoped people wouldn’t notice the heart was gone.

What Explaining Really Signals

So here’s a simple rule: if you find yourself explaining something over and over again, stop. Not because the audience doesn’t get it. But because you didn’t deliver it in a way that makes them feel it.

Explaining is what you do when you missed the emotional moment. When you launched too soon. Or zigged when you should’ve listened.

We’ve all done it. In life. In work. In relationships. We thought the plan made sense, and when it didn’t land, we doubled down on the logic.

But feelings aren’t logical. They’re lived. And if your audience says, “This doesn’t feel right,” you don’t get to talk them out of it. You have to sit in it with them. You have to feel the loss with them. You have to say, “Yeah, we see that. And we’re sorry we didn’t sooner.”

That’s the part Cracker Barrel missed. And that’s the part that makes all the difference.

Belonging Doesn’t Come with a Memo

At the end of the day, people don’t just want clarity. They want connection. They don’t want an explanation. They want to feel understood.

And when something truly resonates, it doesn’t need to be explained. It just lands. It fits. It feels like home.

Cracker Barrel was home for a lot of people. It still could be. But only if they remember that home isn’t something you rebrand. It’s something you honor.

So the next time a brand—or a leader, or a team—decides to make a change, ask the real question:

Does this still feel like us?

And if the answer’s fuzzy? Stop the rollout. Forget the announcement. Sit down and listen.

Because once you’re explaining, you’ve already missed the moment.

And getting it back takes more than just better words. It takes being human again.

Every few months, a new tool promises to change the way we work. It arrives with slick demos, breathless praise, and a promise that this—finally—will be the one to save us from Slack fatigue, calendar gridlock, and the slow death of email. Knowledge work, we’re told, just needs better tooling. Smarter software. A faster shortcut.

And yet here we are—still drowning in tabs, still late to meetings about meetings, still not doing the actual work.

The modern knowledge worker has become a glorified app wrangler.

We’ve traded depth for dashboards. Substance for syncs. Focus for Figma.

It’s not that these tools are bad. Quite the opposite—many are brilliant. But brilliance in design doesn’t equal value in practice. Somewhere between Notion templates and Asana updates, we started optimizing for collaboration theater rather than meaningful output.

The Tyranny of the Workflow

Let’s talk about time.

A McKinsey study found that the average knowledge worker spends nearly 60% of their week on communication and coordination. That’s emails, chats, calls, pings, project updates—the connective tissue of work. Only 40% goes to actual creation.

But here’s the kicker: as more tools promise to reduce this burden, the burden grows. Why? Because every tool demands upkeep. Every platform introduces a new meta-layer of labor: configuring, syncing, tagging, sorting. The tool becomes another inbox, another thing to check.

In theory, Monday.com should save you time. In practice, you now spend Monday updating Monday.

Productivity as Performance

There’s also a deeper, more psychological shift at play. The modern workplace is obsessed with visibility. In remote and hybrid environments, presence has become proxy for productivity. We perform busyness because it’s safer than being quiet. A flurry of Slack messages at 9 p.m. is the new badge of commitment.

Productivity tools enable this performance beautifully. You can create stunning dashboards that show progress, even when no real progress is happening. You can comment, tag, and emoji-react your way into the illusion of momentum.

It’s the professional version of jogging in place.

The Cost of Context Switching

Every tool you add fractures your attention. One minute you’re in Miro brainstorming, the next you’re in Jira filing tickets, then over to Loom to record a recap, then back to Slack to react to the recap. The task isn’t the task anymore—managing the tools has become the job.

Cognitive overhead piles up. Studies have shown that it can take over 20 minutes to regain focus after a context switch. Multiply that by the dozen apps you toggle through each day, and you start to see the real cost of “productivity.”

We used to open a document and write. Now we open five tools to decide how we’re going to write, where we’re going to write, and who we should tag before we write.

The Real Productivity Hack? Saying No

Here’s a radical idea: stop adding tools.

Most teams don’t need more software—they need more clarity. Fewer silos. Less duplication. More trust in people to work without being constantly monitored or managed by algorithm.

Some of the most productive teams I’ve seen use Google Docs, Zoom, and a shared calendar. That’s it. No supercharged stack. No AI co-pilots. Just a shared understanding of what matters, and the discipline to focus on it.

Because productivity doesn’t come from tools. It comes from priorities.

Let Tools Serve You, Not the Other Way Around

This isn’t an anti-tech rant. It’s a call for discernment.

Use tools where they reduce friction. Kill them where they add it. Default to simplicity. Resist the pressure to be seen as productive, and focus on actually being effective.

Because the truth is, no tool will save us from bad management, fuzzy goals, or the fear of deep work. And the more we chase silver bullets, the further we get from the real target: doing great work that matters.

Sometimes, the most productive thing you can do is close the tab.

A tech company built a hotel check-in system. They stored a million passports and driver’s licenses in the cloud. Then they left the door wide open — no password, no protection, publicly accessible to anyone who knew where to look.

They’re calling it a misconfiguration.

That word is doing a lot of heavy lifting right now.

“Misconfiguration” Is the New “We Got Caught”

The hospitality industry has spent a decade telling guests that digital check-in is the future. Faster. Smoother. More convenient. Scan your ID, skip the line, get to your room.

What they didn’t mention: your passport image now lives in a cloud bucket managed by a vendor you’ve never heard of, governed by security practices you’ll never see, and apparently protected by nothing at all.

A misconfiguration isn’t a glitch. It’s a choice that nobody reviewed. It means someone built a system to handle some of the most sensitive documents a person carries — and the security was an afterthought.

That’s not an accident. That’s a priority problem.

The Real Scandal Isn’t the Breach

The tech industry has a standard playbook for moments like this. Apologize quickly. Use passive voice. Promise a review. Wait for the news cycle to move on.

Nobody asks the harder question: why was a third-party vendor holding a million government IDs in the first place? Why did the hotels outsource this without auditing what happened to the data? Why does “digital transformation” consistently mean transferring risk to customers while transferring revenue to vendors?

You handed over your passport to check into a hotel. You assumed someone responsible was on the other end. You were wrong — and the company responsible will face no consequences that actually hurt.

The hospitality industry sold you convenience. What you bought was exposure.

One million people didn’t consent to their identity documents being public. They consented to checking in faster.

Those are not the same thing. And the industry is counting on you not noticing the difference.

The Trump Mobile T1 is finally shipping. Nine months late. Possibly a rebranded 2024 device. Sold at a premium to people who wanted to believe they were buying something revolutionary.

Nobody ran an algorithm to make that sale. A human did.

Branding Is Not a Feature. It’s a Judgment Call.

The Trump T1 story is not really about a phone. It is about what humans can do that no AI, no machine learning model, and no language processor can replicate: manufacture meaning out of thin air and convince other humans to pay for it.

Apple does not sell smartphones. It sells the feeling of being a certain kind of person. Harley-Davidson does not sell motorcycles. It sells an identity. And the Trump T1 does not sell hardware. It sells belonging to a tribe, a movement, a moment.

That is a profoundly human skill. It requires reading culture, understanding emotion, knowing what people are afraid of and what they are proud of. AI can identify brand sentiment in a dataset. It cannot feel the room. It cannot sense that a particular group of Americans wants something that feels like theirs, something that rejects the mainstream, something that signals defiance.

The Trump T1, delayed, possibly derivative, and sold at a markup, is a masterclass in human brand judgment. The phone’s specs barely matter. The story around it is everything.

Delay Is Not Always Failure. Humans Know the Difference.

Nine months late. In the tech world, that sounds like disaster. Product roadmaps are sacred. Launch windows are everything. Miss one, and you lose the market.

But the Trump T1 did not lose its market. Its buyers were not early adopters hunting the latest chipset. They were loyalists who were going to buy it regardless. The delay barely registered as a problem because the humans running the campaign understood their audience.

That is contextual judgment. That is knowing which rules apply to you and which ones do not. AI would have flagged the nine-month delay as a critical risk factor. It would have modeled churn, forecast lost revenue, recommended a crisis communication strategy. It would have been technically correct and completely wrong about what actually mattered.

Human decision-makers at the top of that organization understood something no model could tell them: their customers were not going anywhere. The delay was irrelevant. The relationship was the product.

Imitation Is a Strategy, Not a Scandal

Reports suggest the T1 may be a modified version of an existing 2024 phone. The tech press jumped on this as a gotcha. But business history is full of brands that won by curating rather than inventing.

Kirkland Signature does not manufacture most of what it sells. It selects, packages, and brands products made by others. It is one of the most trusted consumer brands in America. Store brands across every retail category operate on the same principle. The skill is not in the factory. It is in the selection, the positioning, and the promise.

The humans behind the Trump T1 made a judgment call: find a device that works, put it in a package that resonates, and sell it to people who want exactly that thing. Whether or not that was ethically transparent is a separate debate. As a business decision, it reflects a kind of strategic pragmatism that requires human experience, not processing power.

An AI can source suppliers. It cannot decide which shortcuts are worth taking, which compromises will hold, and which ones will eventually blow up. That intuition comes from years of watching decisions play out in the real world.

Here is the uncomfortable part. Human judgment is not always right. The Trump T1 may be a triumph of positioning or it may be a cautionary tale about overpromising. History will decide.

But that risk is inseparable from the skill. The same capacity that lets a human leader read a room, build a brand, and navigate a nine-month crisis without losing their base is the same capacity that leads humans into catastrophic miscalculations. You cannot have one without the other.

AI does not take those risks. It optimizes within parameters set by someone else. It never bets the company on a gut call. It never builds something because it believes in something.

The Trump T1 is a deeply human product. Flawed, late, possibly derivative, and sold entirely on narrative.

That is not a bug.

That is the whole point.

Shopify Is the Landlord and You’re Building on Rented Space

There’s a move every new ecommerce founder makes. They’re proud of it. They post about it. They call it building a business.

They opened a Shopify store. And they have absolutely no idea what that’s costing them.

Not money. Not yet. Something quieter, and in some ways more dangerous.

Their foundation.

The Ownership Nobody’s Talking About

Everyone understands what it means to own property. You bought it. You control it. Nobody can raise your rent, change your terms, or lock you out.

That’s not what you have with Shopify.

What you have is a lease. A very attractive, very functional, very well-designed lease. And like most leases, the terms favor the landlord.

Your storefront. Your checkout. Your customer data. Your entire digital business — built on infrastructure you don’t own, governed by terms you didn’t negotiate, subject to changes you won’t see coming.

Shopify raised prices 30% in 2022. Restructured payment terms in 2023. Merchants didn’t negotiate. They adapted or they left. And leaving, after you’ve built everything on their land, isn’t really leaving. It’s demolishing and starting over.

That’s what a lease is. The landlord sets the terms. The tenant builds a life inside them.

The Illusion Wasn’t an Accident

Shopify gives you real creative control. Your branding. Your aesthetic. Your customer experience. It looks like ownership. It feels like ownership.

It isn’t.

The things that make a business valuable — the systems, the infrastructure, the data — belong to Shopify. Your customer list is technically exportable. But it lives in their environment, under their terms, subject to their decisions.

The moment Shopify changes a policy, your business changes with it. You don’t get a vote. You get an email.

Aesthetic control is not ownership. It’s decoration on someone else’s walls.

First-Time Founders Are the Perfect Target

Experienced operators read platform risk. They’ve been burned. They make the Shopify tradeoff deliberately, with full awareness of what they’re giving up.

First-time founders don’t have that scar tissue. They hear “just start with Shopify” from every blog, every podcast, every mentor — and absorb it as settled wisdom instead of a business decision with real consequences.

So they build. They invest. They pour months into a storefront on rented land and call it their business. By the time the terms shift — and the terms always shift — they’re too deep to move. So they absorb the change, compress their margins, and keep paying rent.

That’s not entrepreneurship. That’s tenancy with a Shopify theme.

The Real Cost Nobody Is Calculating

Every month you build on Shopify, you go deeper into infrastructure you don’t control. Do that for a year, two years, three — and you haven’t built a business. You’ve built equity in someone else’s platform.

The founders who thrive long-term won’t be the ones who used Shopify the most. They’ll be the ones who knew exactly what they were signing when they did.

Right now, most first-time founders don’t read the lease.

That’s the problem.

We spent two decades tightening lightbulbs, swapping out diesel, and fighting about induction stoves—only to let AI bulldoze the whole thing in under five years.

That’s not hyperbole. It’s the inconvenient truth humming beneath the surface of every ChatGPT query, mid-journey image, and TikTok caption now written by a language model. We built leaner, greener infrastructure—then handed the bill to machine learning. And it’s charging it straight to your power grid.

Welcome to the AI-powered climate backslide.

Big Compute, Bigger Appetite

AI doesn’t sip electricity. It gulps.

Each time you marvel at how fast a chatbot writes your email or generates your “yearbook photo” in vintage 1993 glory, somewhere a data center lights up like a Christmas tree. One query to an advanced AI model can use 10 times the power of a standard web search. Multiply that by millions, add model training, and the real-time inference running 24/7—and suddenly the math gets scary.

In places like Virginia, Texas, and Ohio, utilities are scrambling to build new power plants—natural gas, mostly—just to keep up. Not for hospitals or schools. For cloud infrastructure. One utility literally ran out of capacity and told customers they’d have to wait to install solar panels. Why? Because the juice had already been spoken for—by AI.

Congratulations, You’re Subsidizing Big Tech’s Power Bill

Here’s the part they hope you don’t notice: you’re footing the bill.

Utilities aren’t charging Microsoft or Meta the full cost of all that shiny new infrastructure. Instead, they’re spreading it out—through rate hikes, “grid modernization” surcharges, and special fees you’ll never see unless you read 14 pages of fine print. You thought you were paying for greener energy. Turns out, you’re underwriting server farms.

This is how progress gets stolen. One clever trick at a time.

We spent years reducing household energy usage. Smart thermostats. Efficient appliances. Even the cultural shift—remember when leaving the lights on became a moral failure? But now those gains are being quietly erased. Not by lazy consumers. By enterprise AI compute.

The Energy Paradox No One Wants to Talk About

There’s a name for this kind of reversal: Jevons Paradox. Make something more efficient, and people just use more of it. AI is the textbook case. Better models. Faster chips. Lower costs. So we flood the system with more usage—more queries, more images, more training runs.

The result? Total energy consumption rises, even though each individual task feels cleaner, smarter, cooler.

And that’s the catch: AI is not inherently wasteful. But at the scale we’re now deploying it—blindly, relentlessly—it absolutely is. Especially when the infrastructure can’t keep up. Especially when the grid still leans on fossil fuels. Especially when nobody is regulating how much “intelligence” is worth the ecological toll.

The Tech Industry’s Favorite Cop-Out

Of course, the standard reply is, “But AI will help us optimize everything!”

Sure. In theory, AI can help predict energy demand, improve battery storage, design better solar panels, and run factories more efficiently. That’s the promise. But let’s not kid ourselves: right now, the growth in demand is outpacing every efficiency gain it claims to deliver.

The AI industry is like a guy who burns down your house, then offers to help rebuild it—if you’ll just keep paying him to swing the hammer.

What Happens Next (Hint: It’s Expensive)

If nothing changes, we’ll all be stuck on the same treadmill: cleaner energy tech on one side, supercharged AI consumption on the other. It’s like trying to drain a bathtub with the faucet still on.

And what’s worse, we’re doing it without asking the hard questions:

• Should every task really be “AI-enhanced”?

• Who decides what gets prioritized on the grid—hospitals, homes, or servers?

• When the lights flicker during a summer heatwave, will it be because someone asked a chatbot to rewrite their Tinder bio?

The Way Out (Yes, There Is One)

Here’s the good news: we’re still early enough to choose a different path. But it won’t happen by accident.

We need to stop treating AI like magic and start treating it like any other industrial revolution—messy, resource-hungry, and in need of real governance.

That means:

• Forcing utilities to stop passing AI costs to regular consumers.

• Requiring data centers to prove their clean-energy sourcing, not just promise it.

• Investing in power infrastructure that actually keeps pace with demand.

• And yes, maybe even slowing down deployments that can’t justify their environmental cost.

Because here’s the truth we’ve been dodging: not every AI feature is worth the carbon.

This isn’t about being anti-AI. It’s about refusing to let another shiny tool become another slow disaster. We did the hard work of saving energy. We don’t need to let Silicon Valley burn it all for us. Let’s not sacrifice the planet’s progress for a better autocomplete.