I see so many people online questioning their pursuit of hard science or skills like software development in a post-LLM world. I want to share some thoughts on why I believe this concern, while understandable, misses the bigger picture.

First, let's address the elephant in the room: the "AI influencers" aren't incentivized to give you a measured, nuanced, or hopeful take on current advancements. The algorithm loves top 10 lists and rage bait. Rarely does morality, experience, or deep evaluation from experts play any role in what gets amplified.

This creates a distorted view of what's actually happening in the field. The loudest voices aren't necessarily the most informed ones.

If you're an experienced software engineer and you've explored tools like Cursor or Windsurf, you know it's been incredible not to have to write as much tedious glue code. You knew this work was undifferentiated but necessary.

Now, you can spend more time exploring new concepts with an expert-level partner who doesn't get tired of your questions. This can be an incredible accelerator for the motivated individual.

Don't let go of the wheel. Use LLMs to explore, build understanding, and hone your craft.

As underlying models have improved (ChatGPT-4, Anthropic Claude 3.5 Sonnet, ChatGPT-4o, DeepSeek, o1, o1-pro, and now o3-mini), it's been extremely exciting to see problems that were causing friction evaporate over the last 12 months.

But here's the thing: if you've been paying attention, you know that tools like Devin.ai still have a long way to go to be independently capable. Even with agentic advancements. Even with think-time compute models.

This won't always be the case, and that's inevitable. However, as an industry and society, we'd be very shortsighted to think of this as a catalyst to strip away all the junior jobs. That is a bad take.

Instead, use these advancements to:

• Spend less time on boilerplate - Focus energy on the parts that matter
• Work on challenges that build actual business value - Not just keeping the lights on
• Build expertise more quickly - With the incredible resources now at your fingertips

We are currently in a cycle where many of the bigger companies and funds have been belt-tightening, and so there are a lot of candidates in the market. Don't confuse this with the AI-doomer talk track.

I'd look at it as a renewed call to action to think about and build your communication skills.

It has always been the case that the most successful engineers are the ones who can communicate clearly and those who understand that software engineering, and indeed building value in any business, is a team sport.

This is even more true in a post-LLM world. Using these tools to build value and improve your craft requires both curiosity and the ability to articulate intent.

Ideally, you:

• Understand the underlying systems - Don't just copy-paste without comprehension
• Focus on first principles - And the problems to be solved
• Avoid getting lost in minutiae - While still maintaining attention to detail

The future belongs to engineers who can effectively collaborate with AI tools while maintaining their critical thinking skills and deep understanding of systems. These tools amplify your capabilities - they don't replace your judgment.

The motivated individual who combines domain expertise with effective AI tool usage will be incredibly productive. The key is to stay curious, keep learning, and remember that technology is a tool to augment human capability, not replace human insight.

Don't let the noise convince you otherwise. The future of software development is bright for those who adapt thoughtfully and maintain their commitment to excellence.