AI Doesn’t Need to Be Smart to Wreck Tech Careers

replacement

I still see posts claiming “AI won’t replace developers, it makes dumb mistakes.” Yes, it does. Think of the infamous Taco Bell ordering bot that spat out 18,000 drinks. MIT research also shows that AI often fails to boost senior developers’ performance in complex tasks.

But this framing misses the real story. AI doesn’t need to be perfect. It only needs to be cheaper. And that shifts the math of software teams in ways that break the developer job market.

The Contradiction: If AI Is Flawed, Why Are Jobs Vanishing?

A growing body of data points to something unusual:

  • BLS reports: U.S. unemployment is still low (~4.2%), yet IT unemployment rose to 4.6–5.7% in 2025.
  • ADP payrolls: tech roles have shed 200,000+ jobs in 2024–2025, even as broader hiring remains strong.
  • CompTIA trackers: demand is flat or declining for junior devs, while senior openings concentrate in fewer firms.

So how can AI both “fail” at complex reasoning and eliminate thousands of jobs at the same time?

The Secret Lies in Team Math

Let’s model a 14-person engineering team:

  • Roughly 20% of time is research.
  • AI can halve that → a ~10% productivity boost per developer.

That alone compresses the team from 14 → 13 people for the same output.

But here’s the kicker: coordination costs. Team communication overhead scales as:

n(n−1)/2

  • With 14 devs: 91 links
  • With 13 devs: 78 links

That’s 14% less drag just by being smaller. Combine with the 10% productivity bump and you get ~24% effective efficiency gains.

The paradox: AI’s main impact isn’t smarter code, it’s shrinking team size.

The Oversupply

  • 300,000 layoffs/displacements in 2024–25.
  • 120,000 new CS and bootcamp grads annually.

That’s a workforce where supply expands while demand compresses. Everyone applies everywhere. And with AI, even fresh grads can appear “seasoned”, auto-writing resumes, solving LeetCode, ghosting take-home tests.

Hiring pipelines become overwhelmed with noise, and recruiters find it harder to detect real signal.

Juniors vs. Seniors

  • Juniors: Training-ground roles are vanishing because AI does scaffolding, tests, and boilerplate.
  • Seniors: Still needed for architecture and systems thinking, but fewer slots exist. Displaced juniors rebrand themselves as “experienced,” so seniors now face bloated competition for limited growth roles.

It’s not that juniors are uniquely doomed. It’s that they’re the most visible casualties of a shift squeezing everyone.

Outsourcing + AI: A Clearing House of Cheap Labor

  • Outsourcing firms in India, the Philippines, and LATAM are marketing themselves as AI-augmented labor hubs: “One engineer + Copilot = output of three.”
  • They can underbid U.S. salaries at ¼ the cost, while offering firms flexibility and lower fixed payroll.
  • Even outsourcing intermediaries are consolidating, a sign that global developer supply exceeds demand.

For a U.S. CTO facing high interest rates, why hire three juniors when you can pay an outsourcing firm to staff one AI-boosted offshore engineer?

The Macro Headwinds: Interest + Taxes

Here’s where it all converges into a perfect storm:

  1. High Interest Rates

    • Debt financing is expensive. Tech firms can’t just borrow cheap money to carry oversized teams anymore.
    • Investors push for efficiency, not “growth at all costs.”

  2. Tax Amortization Changes

    • Prior to the 2017 U.S. tax reform (Tax Cuts and Jobs Act), companies could immediately expense R&D costs.
    • After the reform (effective in 2022), firms must amortize domestic R&D expenses over 5 years (15 years for international R&D).
    • Translation: instead of writing off a $50M experimental engineering project this year, you only get to expense $10M per year. The other $40M lingers on the balance sheet.
    • This makes it painful to carry large, experimental developer teams, especially those whose output is uncertain or long-horizon.

  3. Combined Effect

    • High borrowing costs + delayed tax relief = punishment for big R&D teams.
    • Firms are incentivized to keep teams small, lean, and AI-augmented.
    • Risky moonshot projects are starved, while incremental, low-headcount work is favored.

When the Price Gets Too High, You Break the Market

Hiring is a market. Like any market, when the price of entry is set too high, you don’t just get “better buyers”, you change who shows up altogether.

Take landlords and tenants.

  • At a reasonable rent, you get stable, middle-class tenants.
  • Raise the rent sky-high, and you don’t suddenly get “super-tenants.” You get an odd mix: people with poor credit, unstable income, or desperate stories. Occasionally you get a brilliant outlier, the scrappy founder who couldn’t get housing elsewhere, but far more often, you get risk.

The same distortion happens in hiring.

  • When companies crank the bar from filtering for the top 10% to filtering for the top 0.01%, they don’t just eliminate mediocrity. And to be fair, given how flooded the market is, they may not feel they have much of a choice.
  • They also eliminate steady, thoughtful engineers, the ones who would build well but can’t grind 500 LeetCode problems or sketch a whiteboard of solutions under pressure.
  • What’s left is a skewed pool: rote memorizers, resume inflators, and AI-assisted “bootleggers.”

History saw this pattern during Prohibition.

  • Alcohol didn’t vanish when it became illegal. It became prohibitively expensive, and demand went underground.
  • The result wasn’t a more sober nation. It was a surge in organized crime, bootlegging, and speakeasies.
  • By raising the “price” of alcohol beyond reason, the system created perverse substitutes, it broke the market.

Now transpose that back to tech.

  • Jobs are the “alcohol.” The hiring bar is the “price.”
  • When the price becomes impossibly high, you don’t get Dijkstra, you lose him. You get bootleg engineers who can game the process.
  • Innovation shrinks. Business formation slows. The very thing companies hoped to buy with their selective filters, quality, gets undermined.

AI and the Masking of Expertise

AI adds another complication: it makes it harder to see who the real experts are. A personal example: I’ve been learning Japanese for years, but my vocabulary is toddler-level. With AI, I can write polished essays in perfect Japanese, masking the brittleness of my knowledge. It looks fine, until the Wi-Fi drops, or a casual lunch conversation reveals how shallow my fluency really is.

The same thing is happening in tech and art. AI can make anyone look seasoned, but the underlying expertise may not be there. Identifying true depth now takes more time and discernment, the opposite of what current hiring practices try to do. Combine that with oversupply and deadlines, and everyone reaches for shortcuts. But shortcuts no longer work, leaving us back in a logjam, and AI compounds both the demand destruction and the oversupply.

Fear, Control, and the Collapse of Creativity

The oversupply also fuels fear-driven management. Companies, knowing they can replace workers easily, impose more control measures: forced return-to-office mandates, overwork culture, and punishments for speaking out. Employers remind staff daily how replaceable they are.

But fear kills creativity. And creativity is the one advantage humans have when competing and collaborating with AI. The very conditions needed to find new business models, trust, slack, psychological safety, are eroded. Without new models, AI is quickly redirected toward chasing the bottom line instead of augmenting human potential. Even Andy Jassy has said openly that companies “won’t need as many people.” Cutting is easy for management; building new business models is risky and can jeopardize a brand if it doesn’t land, though it can also swing fortunes if it does. That tradeoff is far easier for small operators than for multi-nationals. For large corporations, the safer choice is to cut costs and assert control. But when cuts happen at scale, by the dozens, the impact on the labor pool and on morale is outsized.

Summary: The Market Is Broken

  • Total developers: ~3M in the U.S.
  • Displacement: 300K layoffs + 120K new grads annually = labor glut.
  • Hiring pipelines: swamped with AI-padded resumes → more noise than signal.
  • Juniors: training-ground erased.
  • Seniors: squeezed by team compression + resume inflation.
  • Outsourcing firms: promise “AI-at-scale for ¼ cost,” accelerating wage pressure.
  • Macro backdrop: high rates + R&D amortization rules punish firms for carrying large experimental teams.
  • Hiring bar inflation: distorts the market, favoring cheaters and rote memorizers over true innovators.
  • AI masking effect: makes it harder to see who the true experts are, adding friction and compounding the logjam.
  • Fear-driven management: suppresses creativity, the one antidote to AI commoditization, ensuring stagnation.

👉 Net effect: AI doesn’t just speed up coding. It restructures incentives so that large developer teams stop making economic sense. 👉 And without the right talent, there’s no new business model. Without new business models, the only way to bolster earnings is through cuts.

Conclusion: AI doesn’t have to be perfect to change everything. It acts like the wind: even minimal forces, applied repeatedly and at the right cadence, resonate and amplify existing economic pressures. Over time, those vibrations fracture the structure of the tech job market, much like the famous collapse of the Tacoma Narrows Bridge, where steady winds at the wrong frequency shook the span apart.

#TechLaborShift #KShapedRecovery #LowTechHighAgency #AlternativeFutures

← Back Home