“Devin works 24/7, needs no benefits, and doesn’t ask questions. At scale, that’s transformational.”

Anonymous fintech exec

Gino Ferrand, writing today from Santa Fe, New Mexico 🌞

In late July, Goldman Sachs quietly hired a new engineer. No LinkedIn post. No press release. Just an integration sprint, some provisioning, and a shift in the global developer stack.

The engineer’s name? Devin.
No last name.
No headcount allocation either.

Devin is an autonomous AI software agent built to take tickets, write code, test, debug, and ship. Unlike GitHub Copilot, it doesn’t just autocomplete. It executes. It commits. It runs. And Goldman didn’t deploy it as a trial or a toy. They deployed it at scale.

Which raises a bigger question...

If Goldman Sachs is automating the bottom of the dev ladder, who’s still climbing?

The Quiet Displacement Has Begun

We’ve already written about the decline in junior developer roles. Since 2022, job postings for entry-level software engineers have fallen by 38%. But this isn’t just a hiring dip. It is a structural change.

CTOs at banks, insurers, and hedge funds are asking a new kind of question...

Why build an early-career pipeline if the AI already knows Python?

That used to be unthinkable. Junior engineers were the workhorses of modernization, migrating COBOL to Java, writing internal dashboards, fixing tests. They were slow but trainable, green but hungry.

Now? A system like Devin can write tests faster than a bootcamp grad and doesn’t take three weeks to onboard.

Goldman’s experiment signals something deeper. The slow-motion reshaping of engineering orgs is accelerating in high-trust, high-cost environments like finance. The same places that used to hoard Ivy League talent are now shipping Jira tickets to a machine.

The End of the Entry-Level Apprenticeship

For decades, engineering leadership had a simple talent model. Hire juniors. Pair them with seniors. Let them grow.

But that model depends on time, budget, and redundancy. And AI erodes all three.

Why pay for mentorship cycles when the machine can scaffold your project in seconds? Why invest in onboarding when your agent already knows the stack? Why train, when you can fine-tune?

It’s not just Goldman. In 2025, we’ve seen this pattern across the board:

  • A Fortune 500 insurer built an internal AI agent to refactor legacy code, then quietly paused all junior hiring in its tech org.

  • A VC-backed fintech replaced its entire QA intern program with Codex-based test agents.

  • A West Coast bank is using Claude to write and revise regulatory scripts for stress testing, bypassing its grad program entirely.

Build faster with LATAM engineers who get AI

Hiring great developers used to be the bottleneck. Now it's about finding people who can move fast, collaborate across tools, and co-build with AI.

TECLA helps U.S. tech teams hire senior-level, English-proficient engineers from across Latin America at up to 60% less cost, and in your time zone.

Want to see how fast you could scale?

But Here’s What Gets Lost

No AI tool today is capable of abstraction. It doesn’t understand domain nuance. It doesn’t challenge assumptions. It doesn’t mentor anyone.

Devin can code. But it can’t become a staff engineer. It will never lead a design review or re-architect a fragile platform. If you cut off the junior talent pipeline entirely, your team will eventually hollow out. You’ll have no one left to promote.

Worse, you’ll have no one left to debug Devin.

We’ve already seen what happens when orgs over-rotate on automation. Duolingo tried to let AI design the user journey. The result? Confused users, broken trust, and a silent retreat from social media.

Engineering teams should take note.

The real risk isn’t that AI replaces junior devs.
It is that companies forget what juniors are for.

Goldman Might Be Early. But They’re Not Wrong.

Devin is not perfect. But it is good enough to replace a very specific kind of developer: the low-context, task-focused, procedural executor.

And that is a lot of junior work.

The challenge now is to reimagine what a healthy engineering ladder looks like in a world where AI eats the first rung. Because if all that’s left is senior ICs managing agents, we’ve lost the human apprenticeship that made engineering scalable in the first place.

The path forward isn’t banning tools like Devin. It is making sure you’re still building humans alongside them.

More to come...

Gino Ferrand, Founder @ Tecla

Keep Reading

No posts found