I'm fascinated by LLM-generated code. This week, two experiences left me speechless.

• One of my software engineers had an 'aha moment' watching Cursor write code for a new app component through a MCP connection to Figma: "5 days of work in just 10 minutes". It made me reflect on all the HTML integrators I've hired over the years.
• I personally revamped an entire Rails app in just minutes. The database design was already in place, but still -- I wasn't prepared for this level of efficiency.

These experiences, coming just days/weeks after Claude 3.7 Sonnet's release and Cursor's updates, prompted me to think about the future of software jobs. I'm not the first to share some thoughts, but I will be honest: is it the inflection point?

Some thoughts:

1. Software companies will operate with smaller development teams. I can already see this change happening through my own companies.
2. Software stacks will grow exponentially as LLMs produce more code, leading to:
   • (a) Increased complexity in maintenance.
   • (b) Growing demand for high-performance computing. For future ref, $AMD trades at $96,76 today.
   • (c) Greater need for senior software engineers.
3. "Vibe coding" will emerge among younger programmers--if something doesn't work or has bugs, they'll simply rewrite entire components, modules, or applications.
4. Entry-level positions for junior developers without math/science backgrounds may become more competitive as AI matures.
5. As development teams shrink, junior engineers will have more opportunities to launch their own products and companies.

In essence, the future will be product-led. Product teams will face fewer technical constraints as their engineering teams become more efficient and compact. However, I still see significant value in best practices, particularly regarding data management and migrations. I'm curious to see how LLMs handle data schema alterations in production environments!