The Evolution of Software Engineering
Contrary to popular belief, AI won’t eliminate programming jobs—it will create more of them. As Kevin Scott points out, we’ll need “full stack developers” who understand systems from top to bottom more than ever.
The role will shift from manually typing code to a caretaker position—someone who can step in when AI misbehaves and debug complex systems at multiple levels of abstraction. This requires deep technical knowledge and the ability to “punch down a level” when things go wrong.
This transformation is already happening. AI handles the heavy lifting of writing code, while humans focus on architecture, debugging, and system design. The key is that these jobs won’t disappear—they’ll evolve.
Making AI Universal Through Efficiency
For AI to deliver on its promises, it must become more accessible. This means making it:
- More affordable through economies of scale
- More efficient to run on limited power budgets
- Capable of running locally on personal devices
- Sustainable from an environmental perspective
The transformer architecture that powers modern large language models wasn’t just a capability breakthrough—it was an efficiency breakthrough. Microsoft and other companies are actively researching the following disruptive efficiency improvements that could make AI universally accessible.
What’s exciting is how AI might help solve energy problems rather than just consume energy. As Scott suggests, AI could help develop sustainable energy sources, potentially making energy “an order of magnitude cheaper” and more abundant. This could solve seemingly unrelated problems like water scarcity through more affordable desalination.
The Memory Problem
Before AI agents can truly transform our lives, they need better memory. Current AI systems struggle to remember previous interactions or even their own actions. This forces users to rebuild context that should be persistent constantly.
Human memory isn’t perfect, but we have ways to retrieve information through context and association. AI needs similar capabilities—not just database lookups, but an iterative process to find precise details when required.
Once this memory bottleneck is solved, AI agents will be able to:
- Complete multi-step tasks without constant oversight
- Maintain context across multiple sessions
- Work on complex problems while we sleep
- Draft emails and handle routine communications
The potential is enormous, but the technology isn’t quite there yet.
Healthcare: The Ultimate AI Use Case
Perhaps the most promising application of AI is in healthcare. Microsoft’s collaboration with Stanford University’s tumor board demonstrates how AI agents can dramatically reduce the time needed for cancer diagnosis and treatment planning, from weeks to hours.
This isn’t science fiction; it’s happening now. AI is already helping doctors make better decisions faster, and this is just the beginning. As Satya Nadella, Microsoft’s CEO, points out, AI can “improve care and reduce cost” simultaneously—a rare win-win in healthcare.
The real promise of AI in healthcare isn’t replacing doctors but empowering them with better tools and information.
The Human Element Remains Essential
Despite all this automation, Kevin Scott emphasizes that the “blindingly important thing” will be people who are sensitive to human needs and can think about long-term societal benefits.
The analogy from Kevin Weil of OpenAI is perfect: Just as calculators changed but didn’t eliminate the field of rocket science, AI will transform coding without eliminating the need for engineers. We’ll look back at humans writing code the way we now look at doing rocket calculations by hand—as an outdated approach, but one that evolved into something better.
The future isn’t about fewer jobs—it’s about different jobs. It’s about humans focusing on what we do best while AI handles the rest.
I believe we’re heading toward a world where AI agents help us immeasurably in our daily lives. There are hurdles to overcome and guardrails needed, but the benefits are coming. The only thing holding AI back right now is us—and our willingness to reimagine how we work with these powerful new tools.
Frequently Asked Questions
Q: Will AI completely eliminate programming jobs?
No, according to Microsoft’s CTO Kevin Scott, there will likely be more programming jobs, not fewer. The nature of these jobs will change from manually writing code to overseeing AI systems, debugging them when they fail, and understanding complex systems from top to bottom.
Q: What are the most significant technical challenges facing AI agents today?
Memory is currently the biggest bottleneck. AI systems struggle to remember previous interactions or their own actions, forcing users to rebuild context constantly. Other challenges include making AI more energy-efficient, affordable, and capable of handling physical world tasks (embodied AI).
Q: How is AI already being used in healthcare?
AI is already making significant impacts in healthcare, particularly in diagnosis and treatment planning. Microsoft’s collaboration with Stanford University’s tumor board has reduced the time needed for cancer diagnosis and treatment protocols from weeks to hours, demonstrating real-world benefits for patients.
Q: What does the future of AI look like according to tech leaders?
Tech leaders envision AI agents that can reliably take action on our behalf without constant supervision. These agents would handle routine tasks while we sleep, draft communications, and solve problems autonomously. The transition will be from synchronous interaction with AI tools to asynchronous task delegation.
Q: How can we ensure AI benefits everyone?
Making AI universal requires improving efficiency to reduce costs, developing sustainable energy sources to power AI systems, ensuring the technology can run on personal devices, and creating guardrails to prevent misuse. Additionally, we need people who understand both the technology and human needs to guide AI development toward solving meaningful societal problems.
