On the show floor at this year's KubeCon + CloudNativeCon, amidst the excitement surrounding AI, a simple yet profound message emerged: the top priority for startups, especially those building for systems and platform engineers, is to build solutions that work. Everything else—optimization, scalability, even innovation—comes second.
In an on-stage interview with Vertex’s Megan Reynolds, Kubernetes legend Kelsey Hightower, cut to the heart of the matter. When asked about common pitfalls he sees in startups, his response was striking in its simplicity: “Look, the number one thing is that your product actually works. It literally does what you say it does.” It's a poignant reminder that while cutting-edge tech can dazzle, real value lies in solving practical problems reliably and effectively.
When it comes to unfulfilled AI promises, there is no traveler more weary than the systems engineer. Past AI cycles reminded me of bioluminescence—striking in beauty and brilliance, but ultimately insufficient for real work. When a ML algorithm mistakenly identifies a cat as an elephant, it’s actually hilarious. But when your rogue AI infra agent misaligns memory access and sends your eBPF program into a CPU-maxing death spiral, turning “Hello World” into “∇∞≠¥!” – well, that’s probably one of the very few things we can all agree is wrong and evil.
So, to the systems engineers who long ago abandoned the AI prophecies that promised everything but delivered nothing, I declare: it is a new dawn, and the time has come to build! The ancient promises can now be fulfilled. At KubeCon, I met with many bright-eyed, hopeful founders building solutions today that weren’t possible before:
- Incident response systems that go beyond detection to actively resolve issues, using AI to intelligently triage alerts, identify root causes, and automatically remediate problems before they escalate.
- Infrastructure optimization tools that leverage machine learning to analyze complex architectures, identify inefficiencies, and recommend concrete cost-saving measures, turning the tangled web of modern systems into a well-oiled machine.
- Code analysis tools that not only flag potential issues but also learn from past mistakes, using AI to catch bugs before they become production nightmares, making "works on my machine" a relic of the past.
- Security scanning solutions that employ AI to stay one step ahead of emerging threats, adapting to the ever-changing landscape of vulnerabilities and exploits and providing real-time protection against the dangers lurking in the digital shadows.
The dawn of AI in platform engineering is real, but success in this new era depends on discipline. By prioritizing functionality above all else, platform engineers can build tools and systems that not only leverage AI’s immense potential but also deliver tangible, reliable value. As I reflect on KubeCon in the midst of a renewed AI fervor, here are the four essential tips I’d share with founders to help ensure their AI solutions are grounded in practical value:
- Solve real problems first: Start by identifying a pressing problem and focus all efforts on addressing it effectively. Avoid distractions from secondary concerns like scalability or aesthetics until the core functionality is proven.
- Set clear success metrics: Define what “working” means in measurable terms. Whether it’s uptime, cost savings, or error reduction, make sure there’s a clear benchmark for success.
- Test in real-world conditions: Your system doesn’t truly work unless it performs reliably under real-world stress. Simulate actual usage scenarios and refine based on what you learn.
- Refine through feedback: Once your core solution is in place, gather feedback and iterate. Use data to drive improvements and ensure that the system continues to meet evolving needs.
The sun is rising. The shadows are retreating. The question isn't whether to step into this new light, but how to harness its power to build something that truly works.
Originally published on LinkedIn.