Goldman’s New AI Dev Works Alone… 12,000 Humans Might Be Next in Line for the Digital Guillotine

Back in 2023, I was at Six Flags in Atlanta for a buddy’s bachelor party. It was one of those 97-degree Georgia days where the humidity slaps you in the face the second you step out of the car. We’d been in line for the “Superman” coaster for over an hour, and after two chili dogs and a 32 oz. Coke (a combo I now recognize as a cry for help), conversation turned to careers. One of the groomsmen, wiping sweat off his sunglasses, says, “I’m learning to code. Probably get into IT. This AI stuff isn’t going to replace developers anytime soon.”

And at the time? That wasn’t a hot take. ChatGPT had just launched a few months earlier, and yeah, it was making headlines… but mostly for answering trivia questions, helping people cheat on take-home tests, and giving startup founders a new excuse to skip market research (“we asked ChatGPT what the customer wants”). It felt like a clever assistant… fun at parties, decent at dinner reservations, and occasionally helpful if you needed to explain quantum physics to a 7-year-old. If you worked in customer service or were driving a truck, sure, maybe your radar was up. But engineers? Developers? They were still seen as the architects of this AI boom… not the ones getting quietly pushed out by it.

Fast forward just two years. I wake up this morning, check the reports, and there it is: Goldman Sachs is now piloting a fully autonomous AI software engineer named Devin. Not an assistant. Not a copilot. A full-on, code-pushing, bug-squashing AI employee.

Now, before you roll your eyes and assume Devin’s just another overhyped science project stuck in some internal test environment… don’t. According to Goldman’s CIO Marco Argenti, Devin is already working on real production tasks. We’re talking full-stack development… front-end, back-end, testing, deployment, the whole thing. Argenti even said Devin is “just as good as any developer.” (And somewhere in a WeWork right now, a junior dev is slamming their laptop shut and Googling “how to become a prompt engineer.”)

Now, for context: Goldman Sachs has 12,000 software engineers on staff. These are high-salaried employees helping move trillions of dollars annually across markets. This isn’t your average SaaS company trying to cut dev costs… this is one of the largest financial institutions on the planet, with over $53 billion in annual revenue, and they’re now evaluating whether to scale Devin across their entire engineering org.

On the outside it might look like a PR stunt… but Goldman isn’t testing this out for fun. Devin was built by Cognition, a startup founded in late 2023 that’s already raised big money from names like Peter Thiel and Joe Lonsdale. They claim Devin is the world’s first “fully autonomous AI software engineer,” and based on Goldman’s reaction, they might not be bluffing.

So what exactly can Devin do? A lot more than autocomplete. Devin doesn’t give you a helpful suggestion and wait for your input. You assign it a task, and it executes: builds the interface, hooks up the backend, writes tests, migrates legacy code, integrates APIs. Basically, all the tedious stuff junior devs spend weeks on… only Devin doesn’t get distracted, doesn’t take PTO, and doesn’t need a product manager to remind it about deadlines every 48 hours.

And Goldman’s not the only one heading in this direction. Microsoft and Alphabet both report that AI is already responsible for 30% of the code written on some projects. At Salesforce, Marc Benioff said AI is doing 50% of the work at the company. We’re not talking about a glimpse into the distant future… it’s happening right now, on real teams, in real workflows.

Argenti even mentioned that AI tools like Devin are producing 3-4x the productivity of previous-generation tools. And that’s the part that really matters. This goes way past writing code faster… it’s really about changing how teams are structured. If Devin can do the job of three engineers, then the equation isn’t “How do we support our team with AI?” It’s “How many engineers do we actually need now?” And once that question enters the office, everything shifts.

Engineering isn’t the only department feeling the heat. Argenti hinted that roles across the entire organization (from compliance to ops) are just as ripe for automation. And Bloomberg is already projecting up to 200,000 job cuts in banking over the next 3–5 years directly due to AI implementation. So if you're still hanging onto that “AI can’t take my job because I do critical thinking” badge… might wanna start reprinting your resume just in case.

Oh, and let’s talk about the “hybrid workforce” spin Goldman is putting on this. Sounds nice, right? Like a utopia where humans and AIs hold hands and write better code together. But let’s stop the toxic optimism: in corporate speak, “hybrid workforce” almost always means “We’ll keep the top 10%, and the rest of you can train the AI model on your way out.” Unfortunately for everyone (except billionaires) in 2025, this is just the way business is being done now. AI doesn’t ask for equity. It doesn’t negotiate comp packages. It doesn’t quit to go start a Web3 side project. From an efficiency standpoint, it’s the dream hire (assuming it actually works).

Now, this doesn’t mean developers are extinct. But the role is changing fast. The most valuable devs going forward aren’t the ones who can write the cleanest React code or memorize every Python edge case… they’re the ones who can define the problem, architect the system, prompt the AI, review the output, and catch the flaws before the AI ships something that sets off compliance alarms.

So yeah, if your day-to-day is still 70% “Googling how to fix a TypeError,” you might be replaceable. But if you understand the big picture and can guide the machine? That’s leverage. That’s job security. That’s what companies are going to start paying a premium for.

At the time of publishing this article, Stocks.News holds positions in Google as mentioned in the article.