01What are the new Gemini models and how do they work?

Gemini 3.6 Flash is the star of the new Gemini lineup. Google released three models at once: Gemini 3.6 Flash, Gemini 3.5 Flashlight, and Gemini 3.5 Flash Cyber. Each one does a totally different job.

The main idea is that bigger isn't always better. For two years, AI companies chased higher test scores. Google went a different way.

They focused on efficiency and special jobs.

Here is how tokens work. Your words get chopped into tokens. The AI reads and writes in tokens.

Every token counts. Gemini 3.6 Flash gives better answers but uses fewer tokens. If the old model needed 1,000 tokens, the new one might only need 700.

Same answer, sometimes better.

02Main features of Gemini 3.6 Flash, Flashlight, and Cyber

Gemini 3.6 Flash is faster, more reliable, and uses fewer tokens. It handles a million requests without falling over. That matters for companies that run AI at scale.

  • Gemini 3.6 Flash: better answers, fewer tokens, high speed and reliability.
  • Gemini 3.5 Flashlight: small, fast, light. Perfect for boring tasks like reading PDFs, sorting emails, and making yes/no decisions.
  • Gemini 3.5 Flash Cyber: a defensive security model. It reads code, finds weak spots, and suggests patches.

Cyber can do static code analysis, bug detection, vulnerability discovery, security recommendations, patch generation, and threat analysis. Think of it as a security analyst that never sleeps.

03How to set up the new Gemini models for your business

Julian Goldie suggests starting small. Pick one boring task you hate and give it to a model. Then add another task, then another.

  1. Choose one boring task you do every week, like sorting messages or summarizing calls.
  2. Give that task to Gemini 3.5 Flashlight. Write a simple prompt that tells the model exactly what to do.
  3. Test it. Make sure it works the way you want.
  4. Once it works, add a second task. Then a third.
  5. Set up agents to run automatically, like a research agent every morning or a support agent that never clocks out.

That is how you build an AI system. Not all at once, but one boring job at a time.

04How to use Gemini models with real examples

For Gemini 3.6 Flash, think of running a whole team of agents. Julian shared an example for his community, the AI Profit Boardroom.

  • Research agent: searches the web every morning for new AI tools and writes a summary of the three most useful ones.
  • Content agent: takes that summary and turns it into five short posts for business owners.
  • Support agent: reads new questions, finds answers from guides, and drafts friendly replies.
  • Planning agent: looks at what members asked about and suggests new training topics.
  • Quality checker: reviews everything the other agents wrote and fixes anything unclear.

For Flashlight, Julian gave a prompt: 'Read this member message, decide if it's about content, automation, sales, or tech, and reply with just that one word.' One word back. Fast, clean, done.

05Results and comparisons from the episode

Julian compared Gemini 3.6 Flash to older models. It uses fewer tokens for the same job, sometimes better answers. That is a big deal at scale.

He also compared Gemini 3.5 Flashlight to models like Claude Haiku or mini versions from other companies. It is small, fast, and no-frills.

For Cyber, he explained it is not a hacking tool. It is defensive. It helps patch security bugs before bad guys find them.

The bigger point: AI is getting lighter to run. When something gets lighter, people run more of it. More agents, more automation, more stuff running in the background.

06Tips, limits, and mistakes to avoid

Julian warned that one model can't be perfect at everything. That is why Google split Gemini into specialists. Don't use a genius model for boring jobs.

Mistake to avoid: building everything at once. Start with one task, get it working, then add more.

Another warning: AI writes code fast, but fast doesn't mean secure. If AI writes your code, you need AI to check it too.

Don't get distracted by benchmark scores. For real businesses, speed, reliability, and token usage matter more.

07Who the new Gemini models are for and why they matter

These models are for business owners who want to automate boring tasks. You don't need to be a developer to use Flashlight for sorting emails or summarizing calls.

Cyber matters even if you're not a developer. If you have automations that collect member info or move data between tools, Cyber can help keep them safe.

The takeaway: the person who sets up the workflow wins. Everyone has access to the same tools. The difference is who actually builds the system.

Start now. Pick one boring task, give it to a model, and let it work. That is what these new Gemini models are for.