“Frontier AI” is the AI that everyone is excited about. It’s a super-encyclopedia that knows everything about everything and is immensely expensive to train.
When I asked Gemini to explain its concept of itself and to exercise its full creativity, it told me that its concept is “Infinite interconnectedness.” The creative project was a planetarium-like structure made of dichroic glass that responds to the viewer with an ever-changing customized display of knowledge. I call this “Gemini’s Temple of Infinite Interconnectedness.” It’s stunningly beautiful.
Gemini told me that most of its queries are so simple that it’s like using a particle accelerator to pop popcorn.
Frontier AI is dramatically overdesigned for most business uses. “Distilled AI” is a much smaller AI program that can handle at least 90% of corporate needs at a tiny fraction of the cost of frontier AI. Now that open-source AI programs are available from Meta, the Chinese and new U.S. start-ups, the next step will be finding programmers who can distill AI.
That won’t necessarily be easy.
https://www.nytimes.com/2026/08/03/opinion/ai-hype-tech-layoffs.html
I Helped Run Lululemon. Companies Need to Stop Kidding Themselves About A.I.
By Julie Averill, The New York Times, Aug. 3, 2026
…
I know that even A.I. cannot tackle an issue it has never been taught to understand or accurately advise a business it knows nothing about. When we met with the company, it didn’t ask us a single question — not about our data, our systems or how we worked. We eventually said no, but it took longer than it should have. Even a room full of people who knew better wanted its promises to be true.
In the meantime, we continued to run pilots — small test versions of tools inside the business before committing to it. I wanted them to work as much as the next person, but the results were the same almost every time. The demo looked like magic. But the tool wanted clean, connected data and decisions made in consistent, repeatable ways. Ours lived in a dozen systems that did not agree, layered with decades of exceptions and workarounds…
We are getting smarter from each round of pilots. The tools keep improving, and the gap keeps narrowing. But better A.I. tools are never going to close it on their own. What’s left is the part that was always ours: the slow, expensive work of cleaning up the messy data, complex human decisions and tangled systems.
This is how the A.I. revolution is actually going, at company after company… [end quote]
Is AI distillation going to be a new career path for computer programmers?
Gemini says:
While it may not always carry the literal job title “Distillation Engineer,” the core discipline of AI Model Distillation and Optimization Engineering is rapidly emerging as one of the most lucrative and high-demand career specializations in software development.
Titles You Will See on Job Boards
If programmers are looking to specialize in this field, they will find these roles posted under several key titles:
-
AI Systems / Infrastructure Engineer (Focus: Low-latency inference, GPU optimization, model serving)
-
Model Optimization Engineer (Focus: Quantization, pruning, distillation pipelines)
-
Applied ML / SLM Engineer (Focus: Fine-tuning open-source models on domain-specific enterprise data)
-
Edge AI Developer (Focus: Shrinking models to run directly on smartphones, cars, or local hardware without cloud dependencies)
This will be a great career change for programmers who are in danger of being displaced by AI.
Hyperscalers are not facing empty data centers; they are restructuring them. Instead of only building massive centralized megastructures for training runs, they are expanding regional, edge-adjacent data center nodes optimized for fast, continuous inference. Model distillation makes AI practical for everyday enterprise adoption—and that sheer adoption volume is what keeps cloud data centers running at near 100% capacity.
Wendy