A Tale of Two AI Realities: AI Took Away High Paying Tech Jobs… Right?

Meet your new co-worker

The CEO of AWS recently said that firing junior developers because “AI can replace them” is “the dumbest thing I’ve ever heard.”

Is this even controversial?

Depending on who you listen to, you’d think we’re already obsolete.

Every week, another company announces layoffs. Even companies reporting record profits — like Cisco — use AI as the excuse to cut jobs.

It’s convenient.

Not like inflation or tariffs might have something to do with it?

But hey, I’m just a lowly software developer and we all know CEOs of massive tech orgs are exceedingly trustworthy.

Why people (might) say AI replaces jobs

1. The money is there. Billions (nearly a trillion) have poured into AI. Investors need a story, and “AI is replacing jobs” is the easiest story to tell.
2. It’s a scapegoat. Execs can trim bloated orgs from pandemic over-hiring and blame AI instead of their own bad planning.
3. Developers became unlikable. For years we posted “day in the life” TikToks — cold brew, kombucha, and $400K salaries to push a button in React. People love to hate us. So when they hear we’re being replaced, they cheer. Can you blame them?
4. It will. There’s almost no doubt that some jobs are more at risk than others. Will it create as many, more or less jobs than it snatches away? Who the hell knows.

Reality is messier. Maybe more boring too.

Why I don’t believe AI replaces even the worst junior

I use AI every day. Claude, Cursor and ChatGPT.

I pay for these tools because I see their value.

Here’s what people miss:

• AI isn’t autonomous. It can’t join a team, understand the context, sit in meetings, parse designs, make changes across a codebase, open a pull request, accept review feedback, and deploy to production. That’s a human loop — maybe forever.
• The code is excessive. Ask AI for a helper function and you’ll get a novella. Ask it for a data pipeline and you’ll get a brittle mess. The more complex the ask, the more garbage comes back.
• The “80/20” myth is nonsense. People love to say AI does 80% of the work and you just “review” the last 20%. No. Reviewing broken code isn’t 20%. It’s rewriting. I’ve wasted hours chasing AI-generated nonsense before realizing I could’ve written it clean in 15 minutes.
• Mission-critical code is off-limits. Pipelines that ingest millions of rows, need to scale, and can’t afford inaccuracy — AI isn’t writing those. Its tests are weak. Its assumptions are sloppy. Nobody’s risking revenue streams on it.

Here’s the most interesting thing people seem to skip over:

AI companies are hiring developers at outrageous salaries. If OpenAI and Anthropic are fighting over talent and still hiring front end, full stack and back end developers — then why on earth would your no-name company think it has the secret sauce?

What the research says

Cornell and METR recently ran a study on experienced open-source devs. With AI tools, they actually got 19% slower. Not faster — slower. The devs still thought AI had sped them up.

That’s how strong the hype is.

So on one side you have the AWS CEO saying firing juniors is idiotic.

On the other side you have researchers saying AI actually slows down seniors.

And yet… the internet geniuses

Despite literal CEOs of cloud giants and researchers studying top open-source devs, there will still be a chorus of self-proclaimed geniuses on LinkedIn, Twitter, Reddit, you name it, telling us we’ve got it all wrong.

Apparently they’ve figured out how to use AI the “right” way.

Not AWS.

Not Cornell researchers.

Not anyone else.

Just them.

Anything is possible… if you just believe. Or lie.

Is this your reality too?

Recently I was building a data pipeline to ingest hundreds of thousands of TikTok data points. I asked AI to help with some test cases.

At first, the tests looked fine. They were passing. Everything was green. But the actual data flowing through was obviously broken. So I dug deeper.

What did I find? The AI had hard-coded return values. In other words: the tests were designed to always pass, no matter what.

If I hadn’t checked, we could have shipped something that wasted tons of API credits, dumped millions of rows of junk into our system, and left us with a weekend cleanup job.

Obviously, my job is not rubber-stamping AI code.

But ask yourself this — would any junior developer do something this nefarious?

What would you say if you saw your human colleague make this mistake?

That’s not just “oops.” That’s dangerous. And it shows exactly why AI is not ready to be trusted blindly — especially on critical systems.

Don’t take my word for it

If you think AI is already a one-to-one replacement for developers, don’t just argue with me online — test it yourself.

Clone a large open-source project.

Use AI to try and make a meaningful change.

Even better, try to implement a feature or fix a bug and prepare it for a pull request.

(And for the love of God, don’t actually submit the PR.)

If AI was really a total replacement, these projects would already be fully automated. They’re not. And if you try it, I doubt you’ll get very far without running into the same problems the rest of us do.

That little experiment will give you a very clear picture of AI’s actual capabilities — far more than any headline or hype cycle will.

So where does that leave us?

Two realities:

• The hype reality: AI is replacing jobs, productivity is doubled, juniors are toast.
• The ground reality: AI is helpful and messy. Productivity is marginal at best. Firing juniors is dumb.

This is exactly why at Parsity, we’re making a pivot into full stack AI.

Don’t you roll your eyes.

Ignoring the gold rush would be foolish. The real opportunities are in building practical apps: retrieval-augmented generation (RAG), smart search, data summarization and integration. Things teams actually need and what companies are happily paying for.

We want to teach these skills — because those are the ones that will last after the hype fades and they’re honestly not that difficult to learn.

So before you jump into another Twitter fight about AI, try it yourself. Build something non-trivial. Push it until it breaks. Then you’ll see the difference between hype and reality.

We are a small program. We work with everyone one on one. Apply here to learn full stack + AI in a practical way.