The Math Ain't Mathing: AI, Workers, and the Bubble Nobody Wants to Talk About

Design TechniquesCareerNotes from the Studio
Written By Meghan Lewis
Meghan Lewis · August 14, 2026 · ~X min read · This one's long. Grab a coffee — or two.

Last updated: June 3, 2026 — this post is updated as new developments emerge.

Something is happening in the design industry right now that I can't stop thinking about. Junior and mid-level designers and developers are being laid off — not because their work is bad, not because they aren't talented, but because companies have decided that AI can do what they do. I want to talk about why that's wrong. And I want to talk to the people navigating this, because they deserve more than a LinkedIn post telling them to "learn prompt engineering."

A note before you read: I'm not a journalist. I'm a senior designer who has been watching what's happening to tech workers — to people in my industry, at every level — and aggregating what I'm reading and seeing. The sources are real. The pattern they're pointing to is real. And the pattern is not new: companies finding new justifications to extract more from workers while insulating the people at the top from the consequences. The technology changes. The story doesn't.

The Reality

What's actually happening out there.

Companies are cutting junior and mid-level design roles and replacing them with AI tools. The logic, as far as I can tell, goes something like this: AI can generate images, write copy, produce mockups, and iterate quickly. Junior designers do those things too. Therefore, AI can replace junior designers.

This logic is wrong. Not slightly wrong — fundamentally wrong. And the companies acting on it are going to feel the consequences in ways they aren't anticipating yet.

The most recent and staggering example: this week, Meta began laying off approximately 8,000 employees — roughly 10% of its workforce. CNBC reports that junior and mid-level developers are bearing the brunt, competing with AI tools that "work faster and never argue in code reviews."[1] The timing is breathtaking. Meta had just posted record profits of $26.8 billion in Q1 2026. Three weeks later: 8,000 pink slips. The day before the layoffs, Meta's Chief People Officer sent an internal memo announcing that 7,000 employees would be reassigned to new AI-focused teams. Meta also quietly froze or eliminated around 6,000 open positions. So the full picture is: 8,000 people lost their jobs, 7,000 were reassigned to AI roles whether they wanted to be or not, and 6,000 future hires simply ceased to exist. Sources told Reuters that additional layoffs are expected later in 2026, with Meta projected to cut nearly 20% of its total workforce before year end.[2] The reassignment announcement arriving before the layoffs reads as a deliberate PR calculation — a way to frame mass cuts as a "transformation." At the same time, Zuckerberg has been personally recruiting AI researchers with compensation packages reportedly reaching $100 million. The people at the bottom of the org chart are subsidizing the bets being made at the top.

And it isn't just Meta. Snap cut 1,000 jobs — 16% of its workforce — plus 300 open roles eliminated. Disney cut 1,000 positions, mainly in marketing. Block's Jack Dorsey explicitly cited AI when cutting his company nearly in half, from 10,000 to fewer than 6,000 employees. Accenture cut 11,000 roles tied to AI-driven automation. Amazon cut 30,000 corporate jobs across late 2025 and early 2026. In Q1 2026 alone, over 110,000 tech workers were laid off across 137 companies — nearly on pace to match 2023's peak of 260,000. Through mid-May 2026, layoffs.fyi tracks 327 layoff events affecting nearly 139,000 workers — a pace of roughly 1,000 people per day.[3]

Goldman Sachs specifically names graphic designers — alongside management consultants and call center workers — as among the knowledge and creative workers already seeing displacement from AI.[4] That's not a prediction. That's a current-state assessment from one of the world's most-watched economic research teams.

Design isn't engineering, but the pattern is identical — and it's hitting junior and mid-level workers hardest at every level. And it's accelerating.

8,000
Meta workers laid off this week — at a company that had just posted record Q1 profits of $26.8 billion
139,000
tech workers laid off in 2026 so far, tracked by layoffs.fyi — roughly 1,000 people every day
20%
Meta's projected total workforce reduction by end of 2026, per Reuters sources

The signal isn't coming only from economists and industry analysts. In May 2026, three commencement ceremonies turned into unplanned referendums on AI — and the graduates weren't having it. At the University of Central Florida on May 8, commencement speaker Gloria Caulfield, a real estate executive, told the College of Arts and Humanities and Nicholson School of Communication graduates that "the rise of artificial intelligence is the next industrial revolution." Thousands of students erupted in boos. Someone yelled "AI SUCKS." Caulfield turned around, hands out: "Oh, what happened? Okay, I struck a chord." English creative writing graduate Madison Fuentes captured the room: "Why say that in a room full of creatives? I don't think that kids are having a hard time accepting it because we know that AI exists. I think we're just having a hard time acknowledging that it's taking away job opportunities from us." A local news Instagram poll afterward found 88% of respondents wanted AI out of their field.[42] A week later, former Google CEO Eric Schmidt was booed repeatedly at the University of Arizona's commencement. When he said AI "will touch every profession, every classroom, every hospital, every laboratory," the boos intensified. He paused and said "If you'd let me make this point, please —" and was drowned out. SiliconAngle noted it was at least the third such incident in two weeks.[43]

The collision of AI and graduation ceremonies took a different turn at Glendale Community College in Arizona on May 15. This time it wasn't a speaker getting booed — it was the technology itself failing in public. GCC deployed an AI system to read graduate names during commencement, and it skipped hundreds of students as they crossed the stage. Names on the jumbotron didn't match the students walking. The ceremony paused multiple times. College President Tiffany Hernandez addressed the crowd: "We're using a new AI system" — and was immediately met with boos. She initially told students they would not be able to walk again. After sustained backlash from the audience the decision was reversed, and affected graduates were called back with human announcers reading their names instead. One graduate, Grace Reimer, later ran the college's written apology through AI detectors. "It has pulled back that it was written with AI," she said. "Which I find hilarious, but also like, what the heck, because you're again not being sincere."[54]

These are not technophobes refusing progress. These are people walking into a job market that has been explicitly restructured against them, being told by billionaires to be excited about it.

I say this as a senior designer who uses AI regularly in my own workflow. I wrote about how I use it in How I Use AI as a Design Tool (Without Losing My Voice) and AI at Every Level: A Designer's Guide from Graduate to Creative Director. I am not anti-AI. I am anti-misunderstanding what AI actually is and what design actually requires.

The Smokescreen

Sometimes AI isn't the real reason. That doesn't make it better.

Here's where I want to add some honest nuance — because the picture is actually more complicated than "AI is replacing everyone."

A December 2025 survey of 1,000 hiring managers found that 59% admit they emphasize AI in layoff announcements because it "plays better with stakeholders" than admitting financial constraints. Research by Metaintro found that AI actually accounts for only 4.5% of documented job cuts — the rest are driven by budget pressures, revenue uncertainty, and post-pandemic workforce corrections.[5] Harvard Business Review published a piece on exactly this dynamic: "Companies Are Laying Off Workers Because of AI's Potential — Not Its Performance."[6]

Companies are using AI as a more palatable public explanation for old-fashioned cost-cutting. Whether the AI narrative is real or performative, the workers on the receiving end of it can't tell the difference.

This matters for a few reasons. First, it means the scale of genuine AI-driven displacement may be smaller than the headlines suggest — which is worth knowing if you're a junior designer trying to calibrate your anxiety. Second, it means companies are actively choosing to frame financial decisions as technology decisions, which erodes trust and makes it harder for workers to understand what's actually happening to their industry. Third — and this is the part that stings — it doesn't actually change anything for the people who lost their jobs. Whether you were cut because of AI or because of a budget shortfall dressed up as AI, you're still out of work.

The honest read: AI is real, the layoffs are real, and the relationship between them is messier and more cynical than most corporate announcements let on. Which makes the case for understanding both what AI actually can do and what companies are actually doing with it even more important.

The Oracle story is its own case study in what that dynamic looks like in practice. In early 2026, Oracle laid off approximately 30,000 employees — roughly 18% of its workforce. A senior Oracle manager posted publicly on LinkedIn that the cuts were "not performance based." The company had just reported a 95% surge in profit. That same week, Oracle hired its first dedicated CFO since 2014, with a compensation package that included $26 million in stock. TD Cowen estimated the layoffs freed up $8 to $10 billion in cash flow. Oracle declined to comment.[40] The people at the bottom absorbed the cost. The person at the top received the reward. LiveMint reported the 30,000 layoffs were scheduled to be completed by June 15, 2026 — even as Oracle reported record cloud and AI growth for the same period.[83]

The Salesforce story from late 2025 fits the same template. In August 2025, CEO Marc Benioff stated in a video interview that he needed "less heads" because AI tools now handled work that had previously required human teams. He framed it as AI-driven efficiency. By December, reports emerged that Salesforce had experienced quality drops and customer relationship problems after letting experienced employees go — work AI hadn't been able to replicate as cleanly as expected. Salesforce issued a clarification: it was a "strategic rebalancing," not a replacement — roles had been redeployed from support into sales, not eliminated for AI. But the CEO's own framing had created the story.[66] Whether that framing was strategic, optimistic, or just loose talk, the result was the same: a major company told the world it was replacing people with AI, then walked it back, and nobody on the outside could tell which version was true.

The New Normal

For the workers who kept their jobs, a different kind of pressure.

Not everyone got laid off. For the workers who survived the first wave of cuts, a new reality is taking shape — one that's worth understanding before it becomes invisible through familiarity.

At Meta, an employee independently built an internal leaderboard called "Claudeonomics" that tracked AI token consumption across the company's 85,000-plus employees. It awarded titles: "Token Legend" for the highest users, "Cache Wizard" for efficiency. In a single 30-day period, Meta employees collectively consumed 60 trillion tokens. The top individual averaged 281 billion — at the cheapest Claude rate, potentially more than $1.4 million in compute for one person in one month. Neither Mark Zuckerberg nor the CTO ranked in the top 250. The leaderboard was taken down two days after Fortune reported on it.[37]

The leaderboard was unofficial, but the pressure behind it wasn't. Meta's Chief People Officer had already told employees that "AI-driven impact" would be a "core expectation" in performance reviews. In January, the company overhauled its review system to incentivize top performers with bonuses up to 200%. Employees have been setting AI agents to run for hours just to maximize token counts. At Amazon, workers have been gaming internal AI leaderboards — running unnecessary, low-value tasks to inflate scores. Fortune called it "productivity theater." Analysts flagged it as a textbook example of Goodhart's Law: when a measure becomes a target, it stops being a good measure.[38]

The New York Times named the phenomenon "token anxiety" — the fear of being the last person in the room not running a swarm of AI agents around the clock. Tech writer Gergely Orosz put the stakes bluntly: "Inside large tech companies, it's becoming a career risk to not use A.I. at an accelerated pace, regardless of output quality." The leaderboards don't measure output quality. They measure consumption. The Times noted the obvious question nobody seems to be asking: are any of these tokenmaxxers actually producing anything good — or is it "a glimmering tower of tokens, constructed by the competitive and fearful, that will topple as soon as we understand what really makes for useful work?"[38]

And someone may be watching more than the count. In April 2026, Reuters reported that Meta began installing tracking software on U.S. employees' work computers that captures mouse movements, clicks, and keystrokes — feeding the data directly into Meta's AI training pipeline. Employees cannot opt out. The tool is called the Model Capability Initiative. Meta's stated goal: build AI agents capable of performing white-collar tasks independently. The memo framing it told employees they could help the company's models by simply doing their daily work.[39]

In May 2026, SF Standard spoke anonymously with a Meta employee of more than a decade, days before layoffs began rolling out at 4 a.m. on May 20. "This is as anxious and stressed as I have ever been at a job," the employee said. When working from home, they spent their breaks "sort of despondent somewhere in my house." On the keylogging and what it signals: "If you're on a work machine, you are probably being surveilled. The framing that we are using this to train AI to do everyone's job — the sort of unapologetic, 'we're training your replacement, and we're not paying you more for it' approach — is just another signal of how little Meta cares about the humans that it employs." On gaming the system to survive: "I do sometimes talk to our internal chatbot and ask somewhat inane questions just to feel like I'm spending some minutes and spending some tokens. I fear that if I don't have those logged minutes, at some point that will reflect really negatively on me." On May 20, the employee confirmed they had been laid off. "It is a huge financial bummer," they said, "but the rest of it is honestly great. I'm definitely relieved."[65] Their summary of this moment: "Even if we haven't lost our jobs to AI yet, we're being commoditized in advance."

And while companies push remaining workers to tokenmaxx their way through performance reviews, the 2026 Gallup State of the Global Workplace report delivers a verdict on what this is doing to the workforce. Global employee engagement fell to 20% in 2025 — its lowest level since 2020, down from a peak of 23% in 2022. Three consecutive years of decline. No region of the world saw engagement increase last year. Not one. Gallup estimates the disengagement is costing the world economy $10 trillion in lost productivity annually — 9% of global GDP.[67] That's not just "people are unhappy." That's teams delivering less, managers burning out, and workers checking out at the exact moment companies are demanding maximum AI productivity metrics. The quarterly profits from layoffs look good on paper. But laid-off workers don't spend. And AI doesn't either.

The workers who didn't get laid off are being asked to tokenmaxx their way through performance reviews while their keystrokes train the AI that might replace them next.

The Pushback

The public isn't buying it.

The numbers and the noise are one thing. But something else is happening too — quieter, more personal, and arguably more significant: the public is pushing back, and the market is starting to listen.

71%
of Americans oppose an AI data center in their community — higher than nuclear plant opposition (53%) on Gallup's very first ask
50%
of U.S. adults say AI's growing use makes them more concerned than excited (Pew Research, March 2026)
68%
of consumers regularly question whether the content they see is real — 50% prefer brands that avoid generative AI (Gartner)

A Gallup survey conducted in March 2026 found that 71% of Americans oppose having an AI data center built in their community — with nearly half, 48%, strongly opposed. To calibrate how significant that is: Gallup has asked about nuclear power plant opposition since 2001, and the all-time high was 63%. This was Gallup's first year asking about data centers, and the opposition blew past that ceiling on its very first appearance. Only 53% of Americans oppose a nuclear plant in their backyard — nearly 20 points lower than data center opposition.[46]

The New York Times put a headline on what the polling is showing: "Tech Leaders Are All In on A.I. The Public, Not So Much." Published in February 2026, the piece noted that Sam Altman himself had admitted AI was spreading more slowly than he expected. The Times drew the historical comparison plainly: radio was touted as bringing "perpetual peace on earth," television would end war, cable would educate the masses and lead to widespread enlightenment. "This time, though, the masses have not been won over."[47] A Pew Research analysis from March 2026 found that half of U.S. adults say AI's growing use makes them feel more concerned than excited.[48]

I was laid off in 2001. I don't remember "loving" the dot-com boom.

The people being told AI is the next industrial revolution aren't the ones benefiting from it. And increasingly, they know it.

The market is starting to reflect that. In April 2026, the Wall Street Journal reported that brands including Aerie and Le Creuset were running "No AI" pledges as affirmative marketing positions — proactively disclosing AI-free production to get ahead of consumer skepticism. Aerie's Instagram post — "No AI-generated bodies or people. Real people only." — became the brand's most popular post in over a year, earning more than 40,000 likes and 500 comments. The consumer data behind the trend is striking: Gartner found that 68% of consumers frequently wonder whether the content they see is real, and 50% would prefer to give their business to brands that avoid generative AI in their marketing.[49]

Institutions are drawing their own lines. Wikipedia's 260,000 volunteer editors voted 40-to-2 in March 2026 to ban AI-generated article content entirely — citing hallucinations, fabricated sources, and the platform's core commitment to verifiability.[51] Harvard's Faculty of Arts and Sciences announced in April it would phase out ChatGPT Edu after June 2026, citing low student uptake and cost, and switch to Anthropic's Claude Code instead.[52] Spotify launched a "Verified by Spotify" badge in April 2026 to distinguish human artists from AI-generated profiles, as tens of thousands of AI-generated tracks flood the platform daily.[50]

Even OpenAI's core user base has begun to fracture. After OpenAI signed a deal with the Department of Defense in February 2026 — agreeing to let the Pentagon use its technology for "any lawful purpose," including military applications — a "QuitGPT" campaign surged, with users citing concerns about surveillance and lethal AI. The backlash was significant enough that Claude hit number one on the App Store as users switched platforms.[53]

One government has started to respond directly. On May 21, 2026 — the day after Meta announced 8,000 layoffs — California Governor Gavin Newsom signed Executive Order N-6-26, becoming the first governor in the nation to formally direct state agencies to study and mitigate AI's workforce disruptions. The order tasks the Labor and Workforce Development Agency with recommendations within 180 days covering severance policy, WARN Act revisions, subsidized employment, worker ownership models, and how unions are incorporating AI into collective bargaining. Two days earlier, the California Senate passed the No Robo Bosses Act, which prevents AI from being the sole reason a worker is fired or disciplined.[76] The California Labor Federation welcomed the order but made clear it wasn't enough: "Catastrophic job loss from AI is not inevitable. It's a political choice."

The resistance isn't just sentiment. It's showing up in graduation halls, in consumer choices, in how institutions are spending their technology budgets, and in polling that puts AI data centers below nuclear plants on the list of things people want near their homes. The pushback isn't coming. It's here.

The Misunderstanding

AI doesn't design. It outputs. There's a difference.

Here's what AI does well: it generates. Given a prompt, it will produce something. Quickly, tirelessly, without complaint. It can iterate on a direction, produce variations, write headlines, suggest color palettes, draft copy. It is genuinely useful for all of these things.

Here's what AI cannot do: evaluate.

It cannot look at an output and know whether it's right. It cannot feel when something is off. It cannot read a client's hesitation in a presentation and pivot in real time. It cannot carry institutional knowledge about a brand, understand the cultural context of an audience, or make the judgment call that saves a project from going in the wrong direction. It has no taste. It has no stakes.

The stakes of algorithmic output without human evaluation extend into hiring itself — directly affecting the junior designers and developers this post is written for. The largest study of AI hiring algorithms ever conducted, led by researchers at Stanford, Chapman University, and Northeastern University and published in May 2026, analyzed 4 million job applications across 156 employers, all screened by Pymetrics, an AI talent platform that assesses candidates through online games. The paper, "Algorithmic Monocultures in Hiring," documented what the researchers call a systemic rejection problem: 42 identical models were shared across employers, meaning a candidate flagged by the algorithm at one company was likely to be flagged by all of them — across every organization running the same system, without any of them knowing. 26% of Black applicants and 15% of Asian applicants applied to positions where the AI discriminated against their group. If those candidates had been recommended at the same rate as the most-favored group, 40,000 more applications would have advanced. Job seekers needed to apply to at least 25 different positions to be almost certain of a single recommendation to move forward. The companies, the researchers noted, had simply plugged it in without asking whether it worked fairly.[80] AI output without a trained human to evaluate it isn't just bad design. Sometimes it locks people out of the door entirely.

AI output without a trained human eye to evaluate it isn't design. It's unreviewed content waiting to cause a problem.

I'll add a note from my own experience. After months of using AI tools as a senior designer in real workflows, the thing that strikes me most is how much human direction the tools actually require. I have to be the "computer" in the partnership — reminding the assistant of things it should carry forward, noticing when something was omitted that should have been included, asking why a decision was made and getting no reason why, which erodes trust. Prompt engineering isn't a minor skill. It feels like an art form that requires as much craft as the design work itself. The AI will do what you ask, in a linear fashion, and do it well. But the creative vision, the awareness of what's missing, the judgment about what to ask for next — that's still entirely on the human. That's the part that can't be automated. And the companies replacing junior designers with AI tools are eliminating precisely the people who are still developing that judgment.

The prompt is not the work. The prompt is the beginning of the work. What comes after — the critical eye, the iteration, the judgment about what's usable and what isn't, the communication to a client about why — that is design. And that requires a human.

A junior or mid-level designer is not just someone who generates things. They are someone who is building or refining the ability to evaluate. Who is developing the taste and the judgment that takes years to build. That process cannot be automated. And cutting it short doesn't save money — it breaks the pipeline.

The Pipeline Problem

Senior designers don't appear from nowhere.

Every senior designer you've ever hired was once a junior designer. They became senior by doing the work — by being in rooms where decisions got made, by making mistakes on real projects and learning from them, by building relationships with clients and colleagues, by developing the instincts that only come from accumulated experience.

Junior designers are not a cheaper version of senior designers. They are the source of senior designers. If you cut the junior pipeline, you are not just solving a short-term budget problem — you are eliminating your own future talent pool.

The industry will feel this in five years. The studios and companies that gutted their junior ranks in 2024 and 2025 will find themselves unable to hire the mid-level and senior designers they need, because those designers won't exist in the numbers required. You can't fast-track a decade of experience.

The Economics

The math ain't mathing.

Here's the part that should give every executive who cut a junior design role pause: the cost savings they think they're getting may not exist.

1 The Compute Bill

Bryan Catanzaro, Nvidia's VP of Applied Deep Learning, recently told Axios: "For my team, the cost of compute is far beyond the costs of the employees."[7] This is coming from someone at the company that builds the hardware powering the AI gold rush. He's not anti-AI — he's describing the reality of what it actually costs to run it at scale.

Jensen Huang, Nvidia's CEO, put his own version of it into numbers: a software engineer earning $500,000 a year should be spending at least $250,000 worth of AI tokens annually — and he'd be "deeply alarmed" if they weren't.[8] Apply that ratio to a mid-level engineer earning $130,000 and the company is spending more on AI tokens than on the person. The math flips fast.

Uber's CTO, Praveen Neppalli Naga, made it concrete in April 2026: the company had blown through its entire annual AI budget — by April. "I went back to the drawing board," he told The Information, "because the budget I thought I would need has already evaporated."[32] This is one of the world's largest technology companies. Out of AI budget before the year was half over.

2 The Jevons Paradox

As tokens get cheaper per unit, usage explodes. The total bill doesn't shrink — it grows.[9] This is the Jevons paradox: make a technology more efficient and people consume exponentially more of it. One startup CEO publicly shared a $113,000 monthly AI bill as a badge of honor. Some individual developers are spending more on tokens than their own salaries.[10] The "it'll get cheaper" argument is already failing in practice.

3 The Evidence Gap

Despite the narrative, the data doesn't support the idea that AI is actually displacing jobs at scale. According to the Yale Budget Lab, there is no widespread evidence that AI is currently replacing human workers in the numbers being claimed.[11] The companies cutting junior designers aren't responding to a documented productivity shift. They're responding to a story — and paying a premium to do it.

4 The Real Math

The short version: you fired a junior or mid-level designer to save $60,000–$90,000 and are now spending $95,000 a year on tokens for the AI that's supposed to replace them. And you still need someone to evaluate the output. The savings evaporated. The work still needs a human.

5 The Adobe Effect

If you're a designer who uses Adobe Creative Cloud, you're already feeling this firsthand. Adobe has tokenized Generative Fill, Expand, and Background — features that used to feel like part of your subscription now consume generative credits with every use. Early subscribers get 100–250 credits a month. Newer users get 25.[12] That's the direction this is heading: AI features that feel free until they aren't, and usage caps that arrive just as you've built workflows around them.

6 The OpenAI Numbers

No single company illustrates the gap between AI hype and AI reality better than OpenAI. In October 2025, buried deep in Microsoft's SEC filing, equity accounting revealed that OpenAI lost approximately $11.5 billion in a single quarter. Microsoft's 27% stake meant its $3.1 billion net income hit implied those losses directly. For context: OpenAI generated roughly $4.3 billion in revenue for the entire first half of that year.[33]

By April 2026, Fortune reported that OpenAI's own CFO, Sarah Friar, had privately warned company leaders that OpenAI may not be able to cover the multi-year cloud and chip contracts it had already signed — even as industry-wide AI data center spending is projected to hit $660 billion in 2026 alone.[34] Altman and Friar issued a joint statement calling the report "ridiculous." The joint statement confirmed the dispute. Meanwhile, Reuters reported OpenAI is laying groundwork for an IPO targeting a valuation of up to $1 trillion — potentially as early as Q4 2026.[35] A company losing $11.5 billion per quarter, whose own CFO is privately flagging it can't cover its contracts, filing for a trillion-dollar IPO. The dot-com playbook, updated for 2026.

7 The Trap

The most rigorous economic argument came in March 2026 from researchers at the University of Pennsylvania's Wharton School and Boston University. Their paper, "The AI Layoff Trap," built a formal mathematical proof of what many had described in words: workers are also consumers. When a firm automates and lays off employees, the automating firm captures all the cost savings — but bears only 1/N of the demand destruction, where N is the number of competitors. So every rational firm has an incentive to automate first, even though the collective result is an economy that produces everything and sells it to nobody. The conclusion: "At the limit, firms automate their way to boundless productivity and zero demand."[81] The researchers tested every proposed policy solution — UBI, retraining, worker equity, capital income taxes. None were sufficient. The only intervention that could correct the distortion: a Pigouvian automation tax — a charge on replacing human tasks with AI that forces firms to internalize the demand they're destroying before they destroy it.

Companies are cutting human designers to save money while simultaneously pouring that money — and more — into AI compute costs. The savings they think they're making aren't landing where they think they are.

The Real Cost

AI isn't just expensive. It's consuming finite resources.

There's a cost to AI that doesn't show up on anyone's subscription bill — and it's arguably the most consequential one.

✦ Energy

AI data centers are consuming energy at a scale that's genuinely difficult to comprehend. One hyperscale data center can use as much energy as 2 million U.S. households.[13] U.S. data center energy demand is projected to nearly double between 2025 and 2028 — from 80 to 150 gigawatts. That's the equivalent of adding a country with the energy needs of Spain in three years.[14] Microsoft reported that its total emissions have grown 23.4% since 2020, citing AI and cloud expansion as the primary drivers.

✦ Water

Water is the quieter crisis. AI data centers require enormous amounts of water for cooling. The water footprint of AI systems globally could reach between 312 billion and 764 billion liters — in the range of the world's entire annual consumption of bottled water.[15] Many of these data centers are being sited in regions that are already facing water scarcity. One quarter of existing data centers may face more water-scarcity days by 2050.

✦ Communities

Communities are feeling it directly. In Indianapolis, residents mobilized against a Google data center for months before a city-council attorney announced the project wouldn't move forward. Between March and June 2025 alone, community opposition led to $98 billion worth of data center projects being blocked or delayed across the country.[16] In New Mexico, a proposed gas pipeline to power a massive data center project called "Project Jupiter" was rejected by the state land commissioner after significant public opposition.[17]

And then there's Lake Tahoe. In May 2026 — this month — Fortune reported that nearly 49,000 residents on the California side of Lake Tahoe are facing the loss of 75% of their electricity supply. NV Energy, the Nevada utility that has supplied their power for decades, cited data centers in the Tahoe-Reno Industrial Center area as part of the reason for ending service. Their local supplier now has roughly a year to find a replacement — competing against PG&E, Southern California Edison, mining companies, and the data centers themselves.[18] As Danielle Hughes, CEO of the nonprofit Tahoe Spark, put it: "We're 49,000 customers. We have no leverage." This is happening in a high wildfire risk area, where reliable power isn't a convenience — it's a safety issue.

And then there's Utah. On May 4, 2026, over 1,000 Box Elder County residents packed a commission meeting to oppose the "Stratos Project" — a 40,000-acre AI data center backed by Shark Tank investor Kevin O'Leary that would be 2.5 times the size of Manhattan, sited near the already-crisis-level Great Salt Lake.[22] The three commissioners went into a back room and voted unanimously to approve it. No public comment was allowed. As one resident present described it: "Officials were unwilling to engage with a single argument from any of the constituents."[23] Once fully built, the facility is projected to consume nearly 9 gigawatts of electricity — close to what New York City uses at peak summer demand. The opposition is fighting for a November referendum. The resistance is real: active efforts to enact moratoriums or bans on data center development have risen from 8 in May 2025 to 78 today, and the opposition is bipartisan — 55% Republican, 45% Democrat.[24]

And while communities fight data centers at the zoning board, one of the most direct harms is playing out in South Memphis, Tennessee. Elon Musk's xAI has been operating dozens of unpermitted methane gas turbines at its "Colossus" supercomputer facility — pumping formaldehyde, nitrogen oxides, and smog-forming pollutants into a predominantly Black community that already receives an "F" from the American Lung Association for ozone pollution and where residents face cancer risks at four times the national average. The Southern Environmental Law Center filed a formal notice of intent to sue on behalf of the NAACP, alleging multiple violations of the Clean Air Act. xAI operated the turbines for months without permits under a loophole that allows unpermitted operation for up to 364 days. "All too often, big corporations like xAI treat our communities and families like obstacles to be pushed aside," said NAACP President Derrick Johnson.[60]

The broader buildout picture reflects the same resistance. TechRadar reported in 2026 that nearly half of U.S. data centers planned for the year had been canceled or delayed — driven by community opposition, power grid constraints, and supply chain disruptions — and that conditions could worsen.[61]

The legal override took a particularly clear form in Saline Township, Michigan — a rural farming community of under 7,000 people with an annual budget of roughly $1 million. In September 2025, both the planning commission and township board voted to deny rezoning of 575 acres of farmland for a $16 billion OpenAI and Oracle Stargate data center campus. Two days later, the developer, Related Digital, filed suit under an "exclusionary zoning" theory — arguing that since the township had no industrially zoned land at all, it had illegally excluded a "necessary" use. Township officials knew the math: even if they won, they'd been advised the project could likely proceed via other avenues regardless. They settled. The data center is under construction. Saline Township secured roughly $14 million in community benefits — less than 0.1% of the project's value. The township attorney: "Even if we'd won the lawsuit, there was a possibility a data center could have been there anyways."[79] At least 19 Michigan municipalities enacted data center moratoriums in the aftermath. On June 1, 2026, Sam Altman visited the construction site with Governor Whitmer to celebrate what he called "a huge bet on the future of AI."

✦ Oregon

Oregon is closer to home than any of these stories — and it's a microcosm of everything happening everywhere else, with both the victories and the failures fully on display.

Start with the wins. A grassroots coalition led by 1000 Friends of Oregon successfully defeated a controversial proposal that would have rezoned up to 2,000 acres of fertile farmland in Washington County for data center development, defeating the urban growth boundary expansion despite significant political pressure.[73] And in May 2026, the Oregon PUC implemented a landmark rule under the state's POWER Act, requiring data centers consuming 20 megawatts or more to pay their actual share of electricity grid costs — ending an arrangement in which residential customers effectively subsidized the industry. Before the law, data center customers paid about 8 cents per kilowatt-hour; residential customers paid more than 20 cents. New rates take effect June 10. "Oregonians should not bear the costs of explosive data center growth," said Oregon PUC Chair Letha Tawney.[71]

The other front of Oregon's story is Hillsboro — the heart of the state's data center corridor. The Oregonian reported that Hillsboro had been quietly awarding enterprise zone tax breaks extending decades into the future, including one agreement through 2051, under a program nominally limited to five years. A rush of 17 applications from 8 entities arrived in early May before the statewide moratorium deadline. Oregon data centers will receive more than $450 million off their 2026 property tax bills — $85 million in Hillsboro alone. The 11 major data centers in Hillsboro each employ an average of 26 full-time workers while receiving $7.6 million each in property tax breaks — roughly $294,000 per job.[68] After The Oregonian's reporting broke, Hillsboro Mayor Beach Pace ordered a halt to new tax break agreements, possibly indefinitely.[69]

At the state legislative level, the one-year moratorium bill passed the budget committee — but with Senator Janeen Sollman, D-Hillsboro, in opposition. I briefly worked alongside Senator Sollman at a former employer while she was serving in the legislature — I remember congratulating her on winning her seat during my time there. Seeing her become a central figure in this fight from the other side is a strange kind of full circle. Per OPB, Sollman said she could not support the moratorium bill after being pressured to remove expanded enterprise zones from her own economic development legislation aimed at Washington County[70] — the same county receiving $85 million in annual data center tax breaks. On May 19, 2026, she lost her reelection bid in the Democratic primary to Myrna Muñoz, a state employee and democratic socialist, in what OPB called the biggest legislative upset of the primary — the first time an incumbent Democrat had lost to a challenger from within their own party since 2018. Opponents during the campaign portrayed Sollman as a corporate shill eager to open farmland for data centers. SEIU, the Oregon Education Association, and the Oregon League of Conservation Voters all backed Muñoz. The margin was roughly 800 votes, with late ballots flipping what had been an early Sollman lead. Muñoz put it plainly: "The public was told these giveaways were about jobs. They are not. They are about giving massive, long-term tax breaks to some of the wealthiest corporations in the world while working families are left to absorb the cost."[70]

And then there's eastern Oregon. A Rolling Stone investigation — conducted with the Food & Environment Reporting Network — documented what happened in Morrow County, where an Amazon data center has been operating since 2011. County commissioner Jim Doherty noticed a cluster of unusual medical conditions among the county's 45,000 residents, all reliant on the Lower Umatilla Basin aquifer as their only water source. Testing found 68 of 70 wells violated the federal limit for nitrates — some reaching 73 parts per million, more than ten times the state limit. Of the first 30 homes Doherty visited, 25 residents had recently had miscarriages and six had lost a kidney. A 60-year-old who had never smoked had his voice box removed for a smoking-related cancer. "The historical precedent here is Flint, Michigan," said Kristin Ostrom of Oregon Rural Action. Amazon denied the connection.[72]

✦ Coachella

In Coachella, California — a city where 97% of residents identify as Latino — the proposed Coachella Valley Technology Campus would span 450 acres of agricultural land. Residents packed a town hall to oppose it. The developer rebranded it a "Municipal Utility Center," language residents felt obscured what it actually was. Campaign finance records showed Stronghold Power Systems, the developer, had previously donated to the former mayor's campaign. One resident asked the pointed question about heat. A University of Cambridge study published in March 2026 — analyzing 20 years of NASA temperature data across 6,000 data centers worldwide — found that data centers raise surrounding land temperatures by an average of 3.6°F, with hyperscale facilities pushing temperatures as high as 16.4°F above ambient. The heat island effect extends up to 6.2 miles from facilities, currently affecting an estimated 343 million people globally.[74] In Coachella, where summer temperatures regularly top 110°F, the city council had no answer for why data centers are being proposed there at all.[75]

✦ The Override

Here's where it gets darker. When community opposition blocks a data center, there's an increasingly common workaround: invoke national security.

In Utah, the Stratos Project was fast-tracked through a body called the Military Industrial Development Authority — a zoning loophole created in 2007 to support Hill Air Force Base — which gave the developers 80% off property taxes and reduced energy taxes from 6% to 0.5%. The same authority that exists to protect a military base was used to override the democratic objections of over a thousand rural residents.[25]

At the federal level, a July 2025 executive order directed agencies to fast-track data center construction by "easing federal regulatory burdens" — explicitly applying to data centers that "protect national security" and streamlining environmental review processes in the process.[26] And as civilian community opposition has blocked or delayed nearly $64 billion in projects, the Army announced plans to host commercial data centers on military bases — an explicit workaround that sidesteps the local zoning fights entirely.[27]

And then the override went a step further. In late May 2026, Wired published more than 1,000 pages of unpublished reports from the Department of Homeland Security, the FBI, and 80 intelligence fusion centers. The documents reveal a new, previously undisclosed domestic threat category: "anti-tech violent extremism." The term does not appear in any public DHS or FBI handbook — it was created without public announcement. Under it, federal agencies are monitoring local town halls, public budget meetings, and environmental protests where residents express concerns about data centers. The Northern Virginia Regional Intelligence Center flagged activities like "photography" and "observing facilities" near data centers as suspicious behavior. The NAACP Legal Defense Fund warned the category is "incredibly unreliable" and encompasses "peaceful data center protests, AI skeptics, and people unhappy with technology." The designation was enabled by National Security Presidential Memorandum 7 (NSPM-7).[78]

The message being sent to communities is clear: this is happening whether you want it or not. Your concerns about water, power, the environment, and democratic process are acknowledged and overruled. For 'Merica.

The AI that companies are using to replace junior designers is running on infrastructure that communities are actively fighting to keep out of their backyards — and losing.

✦ The Supply Chain

The AI buildout has a geography problem most coverage glosses over: 90% of the world's high-end computer chips are made in one place — Taiwan, an island democracy roughly the size of Maryland, 100 miles off the coast of China. A New York Times investigation published in February 2026 found that despite years of classified government briefings, financial incentives, and tariff threats from two administrations, the U.S. tech industry has "stubbornly refused" to shift where it gets most of its chips. Treasury Secretary Scott Bessent stated it plainly at the World Economic Forum in Davos: "The single biggest point of single failure... if that island were blockaded, it would be an economic apocalypse." A confidential report commissioned by the Semiconductor Industry Association estimated that cutting Taiwan's chip supply would trigger the largest economic crisis since the Great Depression — U.S. GDP would drop 11%, twice as severe as 2008. Companies would have chips for only "several months" before their businesses broke down.[58]

The kicker, buried at the end of the Times piece: Nvidia's first "American-made" AI chip, unveiled with great fanfare at a TSMC factory in Phoenix in October 2025, still had to be shipped to Taiwan for final packaging. The chip that was supposed to demonstrate American AI independence isn't finished in America.

And the administration's tariffs — specifically designed to bring semiconductor manufacturing back to U.S. soil — are doing the opposite. The Center for Strategic and International Studies published a brief in May 2026 with a stark conclusion: "Blanket semiconductor and metal tariffs are crippling the buildout of American data center infrastructure." A 100% semiconductor tariff could raise the cost of AI servers by as much as 75%, potentially adding $75–100 billion in additional infrastructure costs over five years.[59] The policies designed to assert American AI dominance are making the infrastructure more expensive to build — and the people who can least absorb that cost are the workers whose jobs were already cut to fund the buildout in the first place.

✦ The Financial System

None of this means AI is going away. But it does mean the story of "AI is free and humans are expensive" is wildly incomplete. The cost is being paid — in compute bills, in token subscriptions, in energy grids, in water tables, in community opposition. It's just not being paid by the people making the decision to cut the junior design roles.

1 The GDP Gap

Goldman Sachs chief economist Jan Hatzius delivered one of the most damning assessments yet in February 2026: AI investment contributed "basically zero" to U.S. GDP growth in 2025. "I think there's a lot of misreporting, actually, of the impact AI investment had on U.S. GDP growth in 2025, and it's much smaller than is often perceived," Hatzius told the Atlantic Council. The reason: most AI equipment is imported from Taiwan and South Korea. The spending shows up as investment but is offset by the import line — it's adding to those countries' GDP, not ours.[36]

2 The Dot-Com Mirror

The comparisons to the dot-com bubble are no longer fringe. "In 2025, the word 'AI' performs the same magic that '.com' once did," Investing.com wrote. "Companies mention artificial intelligence in earnings calls and add billions to their market caps overnight. The dot-com bubble didn't burst because the internet failed — it burst because valuations assumed it would replace gravity."[28] The IMF has weighed in directly, warning that the AI investment bubble could burst in a way comparable to the dot-com crash — with gains not yet realized in the actual economy, just as lofty internet valuations in the late 1990s were often not based on real revenue.[29]

3 The Numbers

The numbers underneath the hype are hard to square. Over $500 billion per year is projected to be spent on AI infrastructure in 2026–2027. U.S. consumer AI revenue is around $12 billion annually. 95% of enterprises report zero measurable ROI from generative AI investments, according to MIT research. The Shiller CAPE ratio — a key market valuation measure — exceeded 40 in late 2025, a level previously seen only before the dot-com crash.[30] The dot-com bubble erased $5 trillion in market value when it burst in 2000. Analysts now identify 2026–2028 as the highest-risk window for a significant AI correction.[31]

4 The Catch

One important difference from the dot-com era: today's AI spending is largely self-funded by profitable tech giants, not debt-driven startups. Some analysts argue this makes a burst less likely. Others argue it makes the eventual correction larger — because the companies placing the bets are also the ones doing the laying off, and they have further to fall.

5 The Banking Risk

The Federal Reserve Bank of Chicago published a formal analysis examining what they called the "tail risk for banks posed by investments in generative AI."[19] A Bank of America survey found that 45% of institutional investors cite an AI bubble as their top financial risk.[20] Banks have increased their concentration of AI-adjacent lending from 9% to 13% of total commitments since 2015. Oliver Wyman analysts put it plainly: "In 2008, banks discovered they owned far more US housing risk than their internal reports suggested. They might soon discover the same about data-center and digital infrastructure risk."[21]

6 The Verdict

The Musk v. OpenAI trial concluded on May 18, 2026, with the jury finding the case barred by the statute of limitations — decided in 90 minutes. OpenAI's attorneys argued that Musk himself pushed for a for-profit entity at various points, tried to gain control of it, and left when he failed. "I was a fool," Musk told the court. "I gave them free funding to create a startup."[44] Musk lost. But the trial put hundreds of pages of private emails and internal documents into the public record — including evidence of what the nonprofit-to-for-profit conversion actually looked like from the inside. The paper trail it surfaced tells its own story.

7 The Round

On March 31, 2026, OpenAI closed the largest private financing deal in Silicon Valley history: $122 billion at a post-money valuation of $852 billion. Amazon committed $50 billion, Nvidia and SoftBank $30 billion each. Notably, $35 billion of Amazon's commitment is contingent on OpenAI completing an IPO or reaching the milestone of artificial general intelligence.[62] A company losing $11.5 billion per quarter, whose CFO has privately flagged it can't cover its contracts, now valued at $852 billion — with a significant portion of its new capital contingent on either going public or achieving AGI. The dot-com playbook, updated for 2026.

8 The Lobby

In Q1 2026, eleven tech companies — including Alphabet, Meta, Microsoft, Anthropic, and OpenAI — spent $20 million on federal lobbying, according to Issue One, a bipartisan political reform group. That's $226,000 per day. Big Tech's lobbying expenditures have nearly doubled since 2020.[63] And layered on top: "Leading the Future," a super PAC backed by Andreessen Horowitz, OpenAI co-founder Greg Brockman, Palantir's Joe Lonsdale, and Perplexity, entered the 2026 midterms with $70 million in cash and a pledge to spend $100 million total. It runs a Democratic affiliate (Think Big) and a Republican arm (American Mission). A competing AI-industry PAC, Public First — aligned with Anthropic — has also entered the race. Together the two PACs had already spent nearly $24 million as of mid-May 2026.[82] Their explicit goal: a federal regulatory framework for AI that preempts and overrides any restrictions placed on the technology by state governments — the same state and local governments whose communities are documented throughout this post. Their first targets in the midterms include candidates who sponsored state AI safety legislation. The industry isn't just lobbying for favorable policy. It's trying to buy the elimination of local and state ability to say no.

9 The Reset

OpenAI has simultaneously begun resetting its spending expectations downward. CNBC reported the company told investors its compute spending target is now around $600 billion by 2030 — scaled back from earlier projections — even as IPO groundwork targeting a $1 trillion valuation continues.[64] Raising record capital while cutting spending targets. Preparing a trillion-dollar IPO while losing $11.5 billion a quarter. The math still ain't mathing.

The bubble may not burst tomorrow. But the dot-com crash didn't announce itself either. Companies that gutted their junior design teams to fund AI subscriptions may find themselves caught on the wrong side of a very familiar-looking correction — having lost both the people and the institutional knowledge they'll need to rebuild.

The Legal Landscape

The courts are starting to catch up.

The foundation of the AI content pipeline has always had a legal problem: the models were trained on data scraped from the internet, without consent, under the contested argument of fair use. That assumption is now being tested in court — and the early results are not going the industry's way.

On March 2, 2026, the U.S. Supreme Court declined to hear Thaler v. Perlmutter, leaving intact a D.C. Circuit ruling that human authorship is "a bedrock requirement of copyright." AI-generated works — created without meaningful human creative contribution — are not eligible for copyright protection under U.S. law. The Copyright Office had already stated that "prompts alone are insufficient to afford a work copyright protection." The ruling closes a door many AI companies were hoping to keep open.[55]

The implications for creative workers are significant. If AI outputs can't be copyrighted, they can't be owned, licensed, or sold in the same way human-created work can. The "replace designers with AI" strategy isn't just ethically questionable — it's legally limited. The work AI produces, standing alone, belongs to no one.

Which may explain, in part, what happened with Sora. On March 24, 2026, OpenAI shut down its AI video generation app just six months after its high-profile launch — and ended a deal in which Disney had agreed to invest $1 billion in OpenAI and license iconic characters for use in Sora-generated content. The deal is canceled. In its entire run, Sora generated $1.4 million in global revenues, compared to $1.9 billion for ChatGPT over the same period.[56] OpenAI's official reason: a pivot toward robotics. The widely reported reality: the product couldn't survive the copyright minefield around the characters it had been built to generate.

The training data issue runs deeper. Getty Images sued Stability AI for using its photos to train AI models without consent, arguing the outputs constitute lookalike products that can confuse end users. The UK High Court found in Getty's favor on trademark infringement — though the copyright infringement claim was dismissed on evidentiary grounds, as Getty hadn't established which specific images were used in training. The broader principle — that training AI on copyrighted material without consent is legally contested — remains unresolved and is working through courts on multiple fronts.[57]

Your work as a designer, illustrator, photographer, or writer is protectable under copyright. AI output, standing alone, is not. That asymmetry is now settled law.

To the Junior Designers

Your instincts are not replaceable. Your growth is not optional.

If you're a junior or mid-level designer or developer reading this while navigating a layoff or a job market that feels hostile — I see you. I was laid off too. After nearly seven years at the same company. It is disorienting and demoralizing in ways that are hard to describe to people who haven't been through it.

Here's what I want you to hear: you were not replaced by AI. You were replaced by a decision-maker who doesn't understand the work. Those are very different things, and conflating them will send you in the wrong direction.

The skills that make you irreplaceable are not the ones AI is threatening. They are:

1

Your critical eye. The ability to look at something — yours or AI's — and know what's wrong with it. This is developed through exposure, practice, and feedback. You can't shortcut it. Keep building it.

2

Your communication. Design is a conversation. With clients, with stakeholders, with users. AI cannot sit in a discovery call, read the room, or explain a creative decision to someone who doesn't understand why it matters. You can.

3

Your context. You carry knowledge about culture, audience, brand, and history that shapes every design decision. AI works from what it has been trained on. You work from what you know and have lived. That specificity matters enormously.

4

Your curiosity. Junior designers ask questions that senior designers have stopped asking because they assume they already know the answer. That fresh perspective catches things. It is valuable. Don't let anyone convince you otherwise.

Learn to use AI tools — genuinely, not reluctantly. Use them to move faster, to explore more, to get unstuck. But use them as a tool you direct, not a replacement for the thinking you're still developing. The designers who will thrive are the ones who can work alongside AI without outsourcing their judgment to it.

To the Companies

What you're actually cutting when you cut junior designers.

I know this post will mostly be read by designers. But on the off chance anyone making these decisions is paying attention:

When you eliminate your junior and mid-level design and development roles, you are not streamlining your operation. You are eliminating the people who catch what AI gets wrong. You are removing the human judgment that turns AI output into usable design. You are breaking the mentorship chain that produces the senior designers you will desperately need in five years. And you are signaling to the design community — loudly — that you don't understand what design is for.

AI is not a designer or a developer. It is a very fast, very capable tool that requires a trained human to direct it, evaluate it, and take responsibility for what it produces. If you don't have those humans, you don't have a design function. You have a content generation pipeline with no quality control.

And if you're specifically betting your creative and technical workflows on OpenAI: in April 2026, the company's CEO of AGI Deployment went on medical leave, its COO was moved into a new role, and its Chief Marketing Officer stepped down — all in the same week, all while the company is prepping for an IPO.[45] That's not a stable vendor relationship. That's organizational flux at the company you've staked your pipeline on.

That will show up in the work. It always does.

The Bottom Line

The human touch isn't a nice-to-have. It's the whole point.

Design exists to communicate something to a human being. It requires human judgment to know if it's working. It requires human relationships to understand what's needed. It requires human accountability to make sure it's right.

AI can help get there faster. It cannot get there alone. And the junior and mid-level designers and developers who are being cut right now are not the problem — they are part of the solution that companies are too short-sighted to see.

There's a quieter signal worth sitting with. While corporations race toward AI maximalism, younger generations are moving in the opposite direction — buying used iPods, returning to film cameras, seeking out technologies that feel human-scaled and deliberate. It gets framed as nostalgia, but it might be something more intentional: people who grew up watching technology extract value from human connection, choosing to opt out where they can. The resistance showing up in graduation halls is also showing up in what people choose to buy.

And then there's Laika. The Portland studio behind Coraline and ParaNorman has spent close to fifteen years building Wildwood — a stop-motion animated fantasy film set in Oregon's forests, releasing October 23, 2026. Fifteen years of handcrafting puppets, building 120 sets, and animating frame by frame, when they could have cut corners at any point. Director and Laika CEO Travis Knight described the film at announcement: "Our movie is a celebration of artistry over algorithms, and of the belief that films made by hand, with enormous care, can still feel bold, surprising, dangerous, and alive."[77] They didn't optimize their way there. The craft was the whole point.

I'll be honest: I'm tired of updating this post. Not because these stories aren't important — they are — but because they keep coming, faster than I can document them. Commissioners vote behind closed doors. Moratoriums expire and applications rush in before the deadline. Community members pack halls for hours, and officials vote unanimously to approve anyway. At some point, it stops feeling like democracy and starts feeling like performance. Like the billionaires are going to do what they want with our resources, and the hearings are just something they sit through on the way to the vote they've already decided.

But Oregon passed the POWER Act. Hillsboro's mayor called a halt. Thousands showed up to the Utah town hall. Graduation halls full of 22-year-olds booed the billionaires telling them to be excited. Wikipedia voted 40-to-2. California signed a first-in-the-nation executive order. Something is being said, even when it's not yet being fully heard. I'm keeping this post. It's a record. Records matter.

If you're a junior or mid-level designer or developer: keep going. Build your eye. Develop your voice. Learn the tools without becoming dependent on them. The industry needs what you bring. Some companies just haven't figured that out yet.

This post is part of a three-part series on AI and design. The companion pieces: How I Use AI as a Design Tool (Without Losing My Voice) and AI at Every Level: A Designer's Guide from Graduate to Creative Director.

Footnotes
  1. CNBC. "Meta Layoffs Starting This Week Stress Harsh AI Reality Inside Zuckerberg's Company." May 18, 2026. cnbc.com — Also: TheStreet, thestreet.com ↑
  2. Reuters via InformationWeek. "Meta plans additional layoffs in second half of 2026, projected to cut nearly 20% of workforce." informationweek.com ↑
  3. Layoffs.fyi. As of May 2026: 327 layoff events, 138,988 workers impacted, 993 people per day. layoffs.fyi — Also: CNBC via InformationWeek 2026 tech layoff tracker: informationweek.com ↑
  4. Goldman Sachs Research. "How Will AI Affect the US Labor Market?" March 2026. goldmansachs.com ↑
  5. Metaintro. "30,000 Jobs Blamed on AI in 2026 — But Is AI Really the Cause?" 2026. metaintro.com ↑
  6. Harvard Business Review. "Companies Are Laying Off Workers Because of AI's Potential — Not Its Performance." January 2026. hbr.org ↑
  7. Catanzaro, B. (Nvidia VP of Applied Deep Learning). Via Axios, reported in Fortune, April 2026. fortune.com ↑
  8. Huang, J. (Nvidia CEO). All-In Podcast, GTC 2026, March 2026. Via R&D World: rdworldonline.com ↑
  9. The Jevons Paradox applied to AI token costs. Via Medium/Newsarticulated: medium.com ↑
  10. Bar-Joseph, A. (Swan AI CEO). Via Yahoo Finance: finance.yahoo.com ↑
  11. Yale Budget Lab. Via Fortune, April 2026: fortune.com ↑
  12. Adobe Generative Credits FAQ. helpx.adobe.com. Also: Fstoppers: fstoppers.com ↑
  13. Food and Water Watch via Source New Mexico, March 2026: sourcenm.com ↑
  14. Bloom Energy report, January 2026. Via Consumer Reports: consumerreports.org ↑
  15. ScienceDirect — "The Carbon and Water Footprints of Data Centers." sciencedirect.com ↑
  16. Consumer Reports, March 2026: consumerreports.org ↑
  17. Source New Mexico — "NM State Land Commissioner Rejects Application for Gas Pipeline to Power Project Jupiter," March 2026: sourcenm.com ↑
  18. Fortune. "It's Like We Don't Exist: Nearly 50,000 Lake Tahoe Residents Face Power Loss as Utility Redirects Lines to Data Centers." May 12, 2026. fortune.com ↑
  19. Federal Reserve Bank of Chicago. "Tail Risk for Banks Posed by Investments in Generative Artificial Intelligence." 2026. chicagofed.org ↑
  20. Bank of America institutional investor survey. Via Federal Reserve Bank of Chicago analysis: chicagofed.org ↑
  21. Oliver Wyman. "How an AI Bubble Burst Could Shake Global Financial Markets." January 2026. oliverwyman.com ↑
  22. CNN Business. "Why Utah Residents Are Protesting a Massive AI Data Center Project Backed by Kevin O'Leary." May 9, 2026. cnn.com — Also: Washington Examiner: washingtonexaminer.com ↑
  23. Peoples Dispatch. "A Massive AI Data Center Transforms Rural Utah into a National Flashpoint." May 9, 2026. peoplesdispatch.org ↑
  24. Deseret News. "What States Have Banned or Paused Data Center Development?" May 16, 2026. deseret.com — Bipartisan opposition data: Introl: introl.com ↑
  25. Peoples Dispatch. MIDA loophole explanation. May 9, 2026. peoplesdispatch.org ↑
  26. The White House. "Accelerating Federal Permitting of Data Center Infrastructure." Executive Order, July 23, 2025. whitehouse.gov — Analysis: MultiState: multistate.us ↑
  27. Task & Purpose. "Army Plans to Host New Commercial Data Centers on at Least 4 Bases." April 7, 2026. taskandpurpose.com ↑
  28. Investing.com. "The Dot-AI Bubble: Why 2026 Feels Like 2000 All Over Again." October 2025. investing.com ↑
  29. Al Jazeera / IMF. "IMF Says AI Investment Bubble Could Burst, Comparable to Dot-Com Bubble." October 14, 2025. aljazeera.com ↑
  30. Nadcab. "AI Bubble: Is It Real? When Will It Burst? Data-Backed 2026 Analysis." April 2026. nadcab.com — MIT research on enterprise ROI cited therein. ↑
  31. IntuitionLabs. "AI Bubble vs. Dot-Com Bubble: A Data-Driven Comparison." March 2026. intuitionlabs.ai ↑
  32. Naga, P.N. (Uber CTO), via The Information, April 2026. Reported in Axios: axios.com — Also: The Information: theinformation.com ↑
  33. The Register. "Microsoft Earnings Suggest $11.5B+ OpenAI Quarterly Loss." October 29, 2025. theregister.com — Also: Windows Central: windowscentral.com ↑
  34. Fortune. "OpenAI CFO at Odds with Sam Altman Over Missed Revenue Target — Even as AI Capex Is Set to Hit $660 Billion." April 28, 2026. fortune.com ↑
  35. Reuters. "Exclusive: OpenAI Lays Groundwork for Juggernaut IPO at Up to $1 Trillion Valuation." 2026. Via Futurism: futurism.com ↑
  36. Goldman Sachs Chief Economist Jan Hatzius, Atlantic Council interview. Reported by Gizmodo, February 23, 2026: gizmodo.com — Also: Tom's Hardware: tomshardware.com ↑
  37. Fortune. "A Meta Employee Created a Dashboard So Coworkers Can Compete to Be the Company's No. 1 AI Token User." April 9, 2026. fortune.com ↑
  38. New York Times. "More! More! More! Tech Workers Max Out Their A.I. Use." March 20, 2026. Print: Section B, Page 1, March 26, 2026. (Subscription required.) — Also: Fortune. "Amazon's Reported Tokenmaxxing Might Gamify AI Usage, Analyst Warns." May 12, 2026. fortune.com ↑
  39. Fortune / Reuters. "Meta Will Start Tracking Employees' Screens and Keystrokes to Train AI Tools." April 21, 2026. fortune.com — Also: TechCrunch: techcrunch.com ↑
  40. Moneywise. "Oracle Gave Its New CFO $26M in Stock After Firing Up to 30,000 Workers." April 11, 2026. moneywise.com ↑
  41. 404 Media. "Students Boo Commencement Speaker After She Calls AI the 'Next Industrial Revolution.'" May 2026. 404media.co — Also: Inc.: inc.com — WKMG: clickorlando.com ↑
  42. NBC News. "Former Google CEO Eric Schmidt Booed During Graduation Speech About AI." May 18, 2026. nbcnews.com — Also: SiliconAngle: siliconangle.com ↑
  43. NPR. "OpenAI's Sam Altman Takes the Stand to Fend Off Elon Musk's Accusations He 'Stole a Charity.'" May 12, 2026. npr.org — Verdict (May 18, 2026): CNN. "Musk Loses Case Against OpenAI." cnn.com ↑
  44. CNBC. "OpenAI's Fidji Simo Takes Medical Leave, Announces Leadership Changes." April 3, 2026. cnbc.com ↑
  45. Gallup. "Americans Oppose AI Data Centers in Their Area." March 2026. gallup.com — Reported by Fortune: fortune.com and Forbes: forbes.com ↑
  46. New York Times. "People Loved the Dot-Com Boom. The A.I. Boom, Not So Much." February 21, 2026. Print: Section BU, Page 1, February 22, 2026. (Subscription required.) ↑
  47. Pew Research Center. "Americans and Artificial Intelligence." March 2026. Via Newsweek: newsweek.com ↑
  48. Wall Street Journal. "Brands Adopt 'No AI' Disclaimers to Stand Out Amid the Slop." April 6, 2026. Via The Cooldown: thecooldown.com — Aerie Instagram data: Brand Vision Insights: brandvm.com — Gartner consumer data via Dynamis LLP: dynamisllp.com ↑
  49. TechCrunch. "Spotify Introduces Verified Artist Badges to Help Distinguish Humans from AI." April 30, 2026. techcrunch.com — Also: CBS News: cbsnews.com ↑
  50. TechCrunch. "Wikipedia Cracks Down on the Use of AI in Article Writing." March 26, 2026. techcrunch.com — Also: Engadget: engadget.com ↑
  51. The Harvard Crimson. "FAS Plans to Grant Access to Anthropic's Claude, Phase Out ChatGPT Edu." April 28, 2026. thecrimson.com ↑
  52. Tom's Guide. "The 'QuitGPT' Movement Gains Steam as OpenAI's Department of War Deal Has Users Saying 'Cancel ChatGPT.'" February 28, 2026. tomsguide.com — Also: Euronews: euronews.com ↑
  53. AZFamily. "AI System Fails During Glendale Community College Graduation Ceremony." May 18, 2026. azfamily.com — Futurism: futurism.com — Newsweek: newsweek.com ↑
  54. CNBC. "U.S. Supreme Court Declines to Hear Dispute Over Copyrights for AI-Generated Material." March 2, 2026. cnbc.com — Baker Donelson analysis: bakerdonelson.com ↑
  55. BBC. "OpenAI Closes Sora Video-Making App and Cancels $1bn Disney Deal." March 24, 2026. bbc.com (via AOL) — Bloomberg: bloomberg.com — Variety: variety.com ↑
  56. Pennington Law. "AI, Art and Global Approaches to Copyright Law: US Supreme Court Declines to Review Thaler v Perlmutter." March 26, 2026. penningtonslaw.com (notes UK High Court ruling in Getty v Stability AI, November 2025) ↑
  57. New York Times. "The Looming Taiwan Chip Disaster That Silicon Valley Has Long Ignored." February 24, 2026. (Subscription required.) Reporter: Tripp Mickle. ↑
  58. CSIS. "The Impact of Tariffs on the AI Data Center Buildout." May 14, 2026. csis.org — Earlier brief: "How Tariffs Could Derail the United States' $3 Trillion AI Buildout": csis.org ↑
  59. Southern Environmental Law Center. "Elon Musk's xAI Facility Is Polluting South Memphis." February 16, 2026. selc.org — NAACP intent to sue: PBS NewsHour: pbs.org — Democracy Now: democracynow.org ↑
  60. TechRadar. "Nearly Half of US Data Centers Planned for 2026 Canceled or Delayed — and Things Could Soon Get Much Worse." 2026. techradar.com ↑
  61. Bloomberg. "OpenAI Valued at $852 Billion After Completing $122 Billion Round." April 1, 2026. bloomberg.com — CNBC: cnbc.com ↑
  62. Fortune / Issue One. "Big Tech Is Spending $226,000 a Day on Lobbying Congress, Advocacy Group Finds." April 23, 2026. fortune.com — Issue One original analysis: issueone.org ↑
  63. CNBC. "OpenAI Resets Spending Expectations, Tells Investors Compute Target Is Around $600 Billion by 2030." 2026. cnbc.com ↑
  64. SF Standard / Pacific Standard Time. "A Meta Employee Gets Real About the Horror of Working There Right Now." Emily Dreyfuss, May 15, 2026 (updated May 20, 2026). sfstandard.com ↑
  65. HRKatha. "Salesforce Reassesses AI-Driven Workforce Strategy After Large-Scale Layoffs." December 29, 2025. hrkatha.com ↑
  66. Gallup. "State of the Global Workplace: 2026 Report." 2026. gallup.com ↑
  67. Hillsboro Herald. "Tax Breaks for Data Centers in Hillsboro Through 2051." May 2026. hillsboroherald.com — Oregonian/OregonLive. "Hillsboro Is Promising Data Center Tax Breaks Into the 2050s." May 2026. Via Yahoo News: yahoo.com ↑
  68. Daily Tidings. "Hillsboro Data Centers Face New Scrutiny Over Millions in Oregon Tax Breaks." May 2026. dailytidings.com — KATU. "Hillsboro Halts New Data Center Tax Break Applications." May 2026. katu.com ↑
  69. OPB / Yahoo News. "Oregon Would Temporarily Bar Expanded Tax Benefits for Data Centers." 2026. yahoo.com — OPB. "What Does the Primary's Biggest Legislative Upset Say About the State of Oregon Politics?" May 29, 2026. opb.org ↑
  70. OPB. "Portland General Electric's Data Center Customers Will Pay More for Electricity Under Landmark Law." May 12, 2026. opb.org — Data Center Dynamics. "Oregon Energy Regulator Approves New Rate Class for Large Load Data Centers." May 2026. datacenterdynamics.com ↑
  71. Futurism / Rolling Stone. "Amazon Data Center Linked to Cluster of Rare Cancers." November 29, 2025. futurism.com ↑
  72. OPB. "Hillsboro Will Not Get More Industrial Land for High-Tech Data Centers." March 3, 2026. opb.org ↑
  73. Fortune / CNN. "Data Centers Are Creating 'Heat Islands' Warming Surrounding Areas by Up to 16 Degrees." March 30–April 1, 2026. fortune.com — CNN: cnn.com — University of Cambridge study (working paper, not yet peer reviewed at time of reporting). ↑
  74. KPBS. "Coachella Residents Call for Data Center Moratorium." April 27, 2026. kpbs.org — NBC Palm Springs: nbcpalmsprings.com ↑
  75. CalMatters. "California AI Order Directs State to Mitigate AI Layoffs." May 21, 2026. calmatters.org — Governor's official EO: gov.ca.gov ↑
  76. Empire Online. "Wildwood Sets October 2026 Release Date as Laika Stop-Motion Fantasy Epic Prepares to Take Flight." January 20, 2026. empireonline.com — Deadline trailer coverage: deadline.com ↑
  77. Wired. "US Law Enforcement Warns of 'Anti-Tech Extremism' as AI Resentment Grows." May 27, 2026. Via AndroidHeadlines: androidheadlines.com — Tom's Hardware: tomshardware.com ↑
  78. Fortune. "A Michigan Farm Town Voted Down Plans for a Giant OpenAI-Oracle Data Center. Weeks Later, Construction Began." May 2026. fortune.com — Tom's Hardware: tomshardware.com — Detroit News: detroitnews.com ↑
  79. Fortune. "AI Hiring Algorithm Has 'Clear Racial Disparities,' Stanford Study Finds." May 26, 2026. fortune.com — The Register: theregister.com — Paper: "Algorithmic Monocultures in Hiring," Bommasani et al., Stanford/Chapman/Northeastern, 2026. ↑
  80. Falk, B.H. & Tsoukalas, G. "The AI Layoff Trap." March 2, 2026. University of Pennsylvania / Boston University. arxiv.org — SSRN: ssrn.com — Reported by The London Economic: thelondoneconomic.com ↑
  81. NOTUS. "AI Industry Super PAC Enters Midterm Elections With a $70 Million War Chest." January 30, 2026. notus.org — NBC News: nbcnews.com — Let's Data Science (competing PACs): letsdatascience.com ↑
  82. LiveMint. "Oracle Layoffs: Tech Giant to Slash 30,000 Jobs to Be Completed by June 15 Despite Booming Cloud/AI Growth." 2026. livemint.com ↑

Further Reading

Go Deeper

  • Investing.com — "The Dot-AI Bubble: Why 2026 Feels Like 2000 All Over Again": investing.com
  • Al Jazeera / IMF — "AI Investment Bubble Could Burst, Comparable to Dot-Com Bubble": aljazeera.com
  • IntuitionLabs — "AI Bubble vs. Dot-Com Bubble: A Data-Driven Comparison": intuitionlabs.ai
  • Gizmodo — "AI Added 'Basically Zero' to US Economic Growth Last Year, Goldman Sachs Says": gizmodo.com
  • The Register — "Microsoft Earnings Suggest $11.5B+ OpenAI Quarterly Loss": theregister.com
  • Fortune — "OpenAI CFO at Odds with Sam Altman Over Missed Revenue Target": fortune.com
  • NPR — "OpenAI's Sam Altman Takes the Stand in Musk Lawsuit": npr.org
  • CNBC — "Meta Layoffs Starting This Week Stress Harsh AI Reality" (May 2026): cnbc.com
  • TheStreet — "Mark Zuckerberg Sends Stunning Message to Meta Employees" (May 2026): thestreet.com
  • InformationWeek — 2026 Tech Company Layoff Tracker: informationweek.com
  • Layoffs.fyi — Live tech layoff tracker: layoffs.fyi
  • Goldman Sachs — "How Will AI Affect the US Labor Market?": goldmansachs.com
  • Harvard Business Review — "Companies Are Laying Off Workers Because of AI's Potential — Not Its Performance": hbr.org
  • Metaintro — "30,000 Jobs Blamed on AI in 2026 — But Is AI Really the Cause?": metaintro.com
  • Fortune — "A Meta Employee Created a Dashboard So Coworkers Can Compete to Be the No. 1 AI Token User": fortune.com
  • New York Times — "More! More! More! Tech Workers Max Out Their A.I. Use" (March 20, 2026, Section B, Page 1 — subscription required)
  • Fortune — "Amazon's Tokenmaxxing Might Gamify AI Usage": fortune.com
  • Fortune — "Meta Will Start Tracking Employees' Screens and Keystrokes to Train AI": fortune.com
  • Moneywise — "Oracle Gave Its New CFO $26M in Stock After Firing Up to 30,000 Workers": moneywise.com
  • 404 Media — "Students Boo Commencement Speaker After She Calls AI the 'Next Industrial Revolution'": 404media.co
  • NBC News — "Former Google CEO Eric Schmidt Booed During Graduation Speech About AI": nbcnews.com
  • CNN — "Why Utah Residents Are Protesting a Massive AI Data Center Backed by Kevin O'Leary": cnn.com
  • Peoples Dispatch — "A Massive AI Data Center Transforms Rural Utah into a National Flashpoint": peoplesdispatch.org
  • Deseret News — "What States Have Banned or Paused Data Center Development?": deseret.com
  • White House Executive Order — "Accelerating Federal Permitting of Data Center Infrastructure" (July 2025): whitehouse.gov
  • Task & Purpose — "Army Plans to Host New Commercial Data Centers on at Least 4 Bases": taskandpurpose.com
  • Fortune — "Nvidia Executive: Cost of AI is Greater Than Cost of Employees": fortune.com
  • Tom's Hardware — "Nvidia Exec Says AI is More Expensive Than Actual Workers": tomshardware.com
  • R&D World — "Jensen Huang Says Spend $250K on AI Tokens Annually": rdworldonline.com
  • Consumer Reports — "AI Data Centers: Big Tech's Impact on Electric Bills, Water, and More": consumerreports.org
  • ScienceDirect — "The Carbon and Water Footprints of Data Centers": sciencedirect.com
  • Source New Mexico — AI data center energy and water concerns: sourcenm.com
  • Fortune — "Nearly 50,000 Lake Tahoe Residents Face Power Loss" (May 2026): fortune.com
  • Federal Reserve Bank of Chicago — "Tail Risk for Banks Posed by Investments in Generative AI": chicagofed.org
  • Oliver Wyman — "How an AI Bubble Burst Could Shake Global Financial Markets": oliverwyman.com
  • Adobe Generative Credits FAQ: helpx.adobe.com
  • CNBC — "OpenAI's Fidji Simo Takes Medical Leave, Announces Leadership Changes": cnbc.com
  • HRKatha — "Salesforce Reassesses AI-Driven Workforce Strategy After Large-Scale Layoffs" (Dec 2025): hrkatha.com
  • Gallup — "State of the Global Workplace 2026 Report": gallup.com
  • OPB — "Portland General Electric's Data Center Customers Will Pay More Under POWER Act" (May 2026): opb.org
  • OPB — "Hillsboro Will Not Get More Industrial Land for High-Tech Data Centers" (March 2026): opb.org
  • Hillsboro Herald — "Tax Breaks for Data Centers in Hillsboro Through 2051": hillsboroherald.com
  • Futurism / Rolling Stone — "Amazon Data Center Linked to Cluster of Rare Cancers" (Nov 2025): futurism.com
  • Fortune — "Data Centers Are Creating Heat Islands, Warming Areas Up to 16 Degrees" (April 2026): fortune.com
  • KPBS — "Coachella Residents Call for Data Center Moratorium" (April 2026): kpbs.org
  • CalMatters — "California Governor Signs First-of-Its-Kind AI Worker Protection Executive Order" (May 21, 2026): calmatters.org
  • Erin Brockovich — AI Data Center community reporting: brockovichdatacenter.com
  • AI Resist List — tracking data center opposition: airesistlist.org
  • Is AI Profitable? — resource tracking AI financial model: isaiprofitable.com
  • Empire Online — "Wildwood Sets October 2026 Release Date" (Laika — "artistry over algorithms"): empireonline.com
  • Wired — "US Law Enforcement Warns of 'Anti-Tech Extremism' as AI Resentment Grows" (May 2026): wired.com
  • Fortune — "A Michigan Farm Town Voted Down Plans for a Giant OpenAI-Oracle Data Center. Weeks Later, Construction Began" (May 2026): fortune.com
  • Fortune — "AI Hiring Algorithm Has 'Clear Racial Disparities,' Stanford Study Finds" (May 2026): fortune.com
  • arXiv / SSRN — "The AI Layoff Trap" (Falk & Tsoukalas, UPenn/Boston University, March 2026): arxiv.org
  • NOTUS — "AI Industry Super PAC Enters Midterm Elections With a $70 Million War Chest": notus.org
  • SF Standard — "A Meta Employee Gets Real About the Horror of Working There Right Now" (May 15/20, 2026): sfstandard.com

Changelog

What's been added and when.

  • June 3, 2026 — Added: FBI/DHS "anti-tech violent extremism" category (Wired, 1000+ leaked pages); Saline Township, Michigan (voted 4-1 to reject, sued, settled — $16B Stargate project proceeding); Stanford "Algorithmic Monocultures in Hiring" study (4M applications, 42 shared models, de facto cross-employer blacklist); "The AI Layoff Trap" economics paper (Falk & Tsoukalas, UPenn/Boston University — mathematical proof of demand destruction); Leading the Future super PAC ($70M war chest, goal of federal preemption of state AI laws); Oracle June 15 layoff deadline; Meghan's personal reflection on prompt engineering as craft. New Economics card 7: The Trap. Lobby card updated.
  • May 26, 2026 — Added: Oregon section (1000 Friends win, POWER Act electricity rate victory, Hillsboro $450M tax breaks and 2050s deals, mayor halt, Senator Sollman vote, Amazon Morrow County cancer/miscarriage cluster); Coachella opposition + Cambridge heat island study (16.4°F, 343M people); Salesforce "I need less heads" / regret walkback; Gallup 2026 employee engagement (20%, $10T lost productivity); California Newsom executive order N-6-26 + No Robo Bosses Act; personal editorial note in The Bottom Line. New resources: airesistlist.org, isaiprofitable.com, brockovichdatacenter.com. Also added: Laika Wildwood — "artistry over algorithms" — to The Bottom Line.
  • May 21, 2026 — Added: The Pushback section (Gallup nuclear vs. data center poll; NYT "not so much"; "No AI" brands/Aerie; Spotify verified badges; Wikipedia ban; Harvard phases out ChatGPT; QuitGPT; Pew approval data); The Legal Landscape section (Supreme Court AI copyright ruling; Sora/Disney shutdown; Getty v. Stability AI); The Supply Chain subsection (Taiwan chip vulnerability; CSIS tariffs crippling buildout); xAI South Memphis pollution/NAACP lawsuit; TechRadar data center cancellations; Musk v. Altman verdict update; OpenAI $122B funding at $852B valuation; Big Tech $226K/day lobbying + $200M super PAC; OpenAI $600B spending reset; Glendale CC graduation AI failure; SF Standard Meta employee firsthand account ("commoditized in advance"). Jump links updated; SF Standard anonymous Meta employee interview added to The New Normal.
  • May 20, 2026 — Added: observer note and jump links; OpenAI financial chaos (CFO vs. Altman, $11.5B quarterly loss, IPO groundwork at $1T valuation); Musk v. Altman "stolen charity" trial; Meta tokenmaxxing, Claudeonomics leaderboard, and keystroke tracking; Oracle layoffs and CFO compensation; UCF and University of Arizona graduation booing; Goldman Sachs "basically zero" GDP contribution finding; New York Times "token anxiety" reporting; Uber CTO AI budget depleted by April; nostalgic tech / authentic living note. New section: The New Normal. Updated: The Smokescreen, The Economics, The Real Cost, To the Companies, The Bottom Line.
  • May 19, 2026 — Initial publication. Sections: The Reality, The Smokescreen, The Misunderstanding, The Pipeline Problem, The Economics, The Real Cost, To the Junior Designers, To the Companies, The Bottom Line.
Meghan Lewis http://www.melewdesign.com
Next

What makes a good designer portfolio (and what doesn't)