Gemini 3.6 Flash and 3.5 Flash-Lite Ignore Temperature, Top-P, and Top-K
Google's Gemini 3.6 Flash and Gemini 3.5 Flash-Lite ignore temperature, top_p, and top_k, according to Google's current migration guide. Existing Gemini requests…
Categories
Only Experts Can Write Good Prompts
Everyone talks about prompt engineering as if memorizing templates and adding "think step by step" will suddenly get you amazing AI results. That's not how it works.
Writing good prompts isn't about knowing special phrases or tricks. It's about having a clear mental model of what you're trying to achieve and being able to translate that model into words. This requires deep expertise in the domain you're working in.
When I prompt an AI to write code I'm translating my mental model of the software into words.
My clients come to me with business problems. They know what they need business-wise. They don't have a mental model of the software they need.
Even if AI could build perfect software unsupervised (it can't), my clients would still need an experienced software developer to understand and translate their actual needs into a prompt.
This is why experienced professionals won't be replaced by AI anytime soon. Their value isn't in writing code or creating documents. It's in knowing what code to write and what documents to create even if they end up not writing the code and the documents themselves but instead a prompt that tells the AI what to write.
Good prompts come from good mental models. Good mental models come from expertise.
What I'm building
Delegate tasks. Get software.
Give Vroni a GitHub issue, bug report, spec, or rough idea. It reads the repo, plans the change, writes code, runs checks, and works toward a review-ready pull request.
Take a look at vroni.com
“Good prompts come from good mental models. Good mental models come from expertise.”
Indeed.
Ethan Mollick just wrote a post about AI producing an “abundance of intelligence”. I beg to differ. AI means an “abundance of information”, which does not automatically translate into an increase in intelligence by the user. Even in the absence of AI, a chasm exists between people who believe they are “smart” versus those who are innately intelligent. Instead of looking at what will change in ten years, we should be looking at what will remain the same–the over 200 known (at present time) cognitive biases impacting human thinking and acting. This in turn impacts the ability to turn any abundance of information produced by AI into meaningful gains by those unable to translate that information into intelligence, or wisdom. As in all human history, base rates, the Pareto principle, and a host of other factors impact the complex adaptive system inherent in any human translating increased information into intelligence.
Just a few (very quick and rough) thoughts.