Samuel Stroschein recently went viral on Twitter, after an acquaintance tweeted about a dramatic move he’d made: laying off his entire engineering team and replacing them with Claude Code.
This kind of layoff is exactly what people fear when they say AI will take our jobs – hence why the post got 2.2M views.
But this layoff wasn’t Samuel’s first choice, and his team even agreed it was the right one. Here’s how AI offered a solution to one of his business’s major bottlenecks to growth.
Why lay off your team for AI?
Samuel is the founder of Opral and originally from Germany, where he built his team. A European HQ made a lot of sense for Opral’s first product, Inlang, which focused on software localization – a day-one problem in the European market.
But Samuel knew that to scale his second product, a change control tool called Lix that the rise of autonomous agents in the enterprise will make critical, he needed to be in the U.S., where bigger customer bases and funding opportunities exist. And moving to New York created a 6-hour time zone gap – leading to serious delays in iteration and R&D cycles.
Samuel started turning to AI tools for instant answers to make progress instead of waiting overnight for human responses.
“Why should I wait 24 hours for my team to reply if AI gives me an instant answer? Now granted, is AI giving me the best answer? No, but it’s still an answer.”
To capture this bigger market opportunity in the U.S., he gave his team an option: Come to the U.S., or I’ll have to let you go. For various personal reasons, they all declined – but they agreed with his choice.
“In my case, I actually got the team onboard. I told them, ‘This is what we have to do’, and they agreed it was right for the product. But this is a very startup-specific scenario where everyone understands a pivot like this.”
How he uses AI as an engineering team
Samuel started out working with Claude Code, but he’s since completely pivoted to Codex, ChatGPT-5’s new coding feature.
“That’s how fast things can change with AI. Codex became so good, it’s the only tool I’m using right now,” he says.
In developing Lix, he “proudly” says 95%+ of his code is now AI-generated. Here’s his process:
1. Ask AI to write the test, and then develop a feature that would pass it
“Once I have the spec in my mind, I tell AI, ‘Don’t implement it - just write the test,’” Samuel says. “That’s a big shift. Building tests becomes so cheap, it’s amazing.”
The tests act as both a blueprint and safety net. And at this phase, Samuel is just looking for something that works.
“Who cares about whether the implementation is the most beautiful or not? If it hits the business requirements, you’re good.”
2. Develop multiple features in parallel
Samuel uses two monitors – one per feature. On each monitor, he runs a separate coding flow with AI, then he bounces between them as the editor.
This essentially lets him context-switch between two projects instantly, because AI maintains the continuity of each thread. He effectively doubles throughput by splitting tasks between AI agents.
“With AI, you can build prototypes much faster,” he says. “The iteration speed is insanely fast.”
3. Let the AI review its own pull requests
Writing code is no longer the hard part – the bottleneck is reviewing and ensuring quality. Samuel’s solution: Use AI to review its own pull requests.
“I have 2-3 different AI agents that review every pull request, and it actually does feel like I have a small team,” he says. “It’s insane.”
Each agent checks the code from a slightly different angle – similar to how human reviewers might catch different issues. This simulates the redundancy and confidence of a team review process but with instant results.
The opportunities and tradeoffs of an AI team
Samuel emphasized that while this approach lowered his burn rate, cost savings weren’t his main driver.
Instead, AI solved a business need that gave Samuel:
- The freedom to get himself in the right rooms. Without time zone issues, Samuel was able to move to San Francisco, embed himself in the U.S. startup ecosystem, and be closer to his customer base.
- The opportunity to capitalize on a moment of need. Samuel says he’s working 5x faster, allowing him to get a product to market while there’s still opportunity.
- The ability to create fast POCs for potential customers and investors. He can now leverage working prototypes instead of hypothetical powerpoint presentations.
But he maintains: “A team with AI is still better than a single person with AI.”
He acknowledges a few tradeoffs that he’s okay with for the time being:
- Settling for a good enough answer. Right now he knows he’s not getting the quality of answers and sleekness of outputs that he would from his human team.
- An AI team still needs a manager. Sometimes his agents don't consider everything when writing and reviewing code, so he still has to steer them in the right direction.
- The negative stigma of the choice. While the framing of the tweet was part of the problem, even after he clarified, he was still painted as the “asshole” for his decision.
So should you replace your engineers with AI?
Take it from the founder who did it: “I would not recommend replacing your team with AI. That could lead to a disastrous outcome. First, make your team faster with AI, and if you have redundancies, act on them.”
This all comes back to a business need. If you’re looking for a reason to replace your team with AI, you probably don’t have one.
Samuel intends to rehire, likely for reviewer and oversight roles. “For now, I’m just going to do the last push myself, leveraging AI heavily, and then once we see adoption, I’ll fill up positions again where I see a bottleneck.”
Here’s his parting wisdom for leaders considering the same choice:
- Prove a business challenge first. If you haven’t tried augmentation or clearly proven there are redundancies in your team, jumping straight to replacement is not a good idea.
- AI-first teams are hard to build. Making AI-first decisions is easier when your team is bought into the vision, but you need to prepare to get past a lot of suspicion and fear.
- You’re going to be the asshole. You might have to make an unpopular decision that’s right for the business - so be empathetic, but own the optics.
