Generative AI is Accelerating Subject Matter Experts

I had time to play with Gemini 3 and Google Antigravity this past weekend, and it took me back to those first “aha!” moments I had when learning programming. This time around, the feeling of awe I get is not from understanding concepts, but rather from supercharging my workflow. That’s where I see generative AI really starting to shine; when used by subject matter experts with clean data and well-guarded rails, large language models like Gemini and ChatGPT drastically cut the time needed to get concepts mapped, MVPs worked up, and even automations launched.

I specify use by SMEs because there absolutely should be a human in the loop; many business use cases require too much nuance (operations) or are too high stakes (finance) to simply hand the wheel to genAI. Often times the data simply isn’t there or isn’t actionable enough to facilitate a full handoff without warm bodies involved. HITL is a symbiotic approach used in machine learning that enhances the capabilities of both humans and machines. The use of said approach by businesses allows their genAI models to be properly trained and ideally more efficient.

The person needs to be well-versed in their business domain in order to maximize the output of AI. The proverbial “what you give is what you get” has never been more true. GenAI loves to sugarcoat things and placate its users, arguably the last thing wanted when building production-ready software or automating financial workflows.

Gemini 3 and Google Antigravity have given me new confidence in where genAI is heading. I was able to get a working prototype generated within 3 prompts/tweaks; what normally would have taken an hour or so to get running took me only 10 minutes. The model’s ability to troubleshoot my local environment and the cloud features I was utilizing far outpaced my previous experiences with online and embedded LLMs.