You can start using AI with a question as ordinary as, "What can I cook with what I've already got?" Type it, say it, or upload photos or screenshots of your ingredients. The voice option called Live accepts images, but currently does not support video or screen sharing. For those, use Advanced Voice in a supported mobile app with an eligible plan. You can use the same account on your phone, tablet, and computer, and reopen your saved conversation as you move between them. Only one Voice conversation can run at a time; features and usage limits depend on your plan and device.
Tell it what's in your refrigerator and pantry, the foods you enjoy, what you avoid, how many people you're feeding, and which pots, pans, and appliances you have. Include how much time you want to spend cooking.
Then ask:
Suggest a meal using these ingredients and this equipment. Give me quantities and clear cooking steps. Ask about anything important I've left out.
For a whole week, give it the meals you want, your budget, and your existing ingredients. Ask for a menu and one grocery list that combines quantities and leaves out what you already have. You check the list and take it shopping.
That's a practical introduction to AI: give it useful information and ask for a result you can use.
I'm Captain Raul Silvas. I use AI for planning, research, writing, everyday organization, and work on our website. Restoring our family sailboat, Valhalla de Tortuga, brings many of those jobs together. The same methods can help with your household, business, hobby, or work.
If you're a Baby Boomer, a Gen Xer, or a Millennial wondering where to start, this series will give you examples you can try.
Make it comfortable to use
I call my main assistant Navigator. We talk through ideas, question assumptions, and turn rough thoughts into a plan.
You can choose an available voice you like, including one with a more masculine or feminine sound. You can also explain how you want the AI to respond:
Be friendly and direct. Explain unfamiliar terms. Give me the short answer first, and tell me when you aren't sure.
Examples and corrections help it follow your preferred style. Supported memory and saved instructions can carry preferences forward when enabled. Voice options and personalization controls vary by experience; OpenAI's voice guide and personalization guide explain the options.
How I direct each member of the crew
Navigator helps me think through the job, then carry out specific tasks. Chat is where we brainstorm and refine the plan. I use Work when the job involves several steps, files, or connected services. An instruction such as "check my email and draft a reply" gives her a task to carry out. Sending requires the appropriate tools and my authorization.
Ongoing inbox checks can also be set up in Work with supported connections and a saved schedule. I specify what deserves attention and how I want to be notified. OpenAI documents that email workflow.
With Raven, the finished result is critical. Raven works with Codex on our website, software, and technical repairs. I get a clearer brief by describing what the finished product must do and how I'll judge it. For example:
Build a small tool that combines daily reports into one weekly summary, grouped by crew, with missing entries flagged.
That gives Raven an output to build and test. I review whether the result actually does the job.
With Jolene, I give her a goal and room to work out how to reach it. An example might be:
Find practical ways to bring more interested readers to our website. Research the options and refine the approach as we learn what works.
Her role allows for investigating alternatives and thinking beyond the first approach. As my cloud-based dot, she can continue assigned work between conversations using her own computer and browser, even while my computer is off. She still needs suitable access and comes back to me when a decision or approval is required. OpenAI explains how dots work.
These are the roles and working methods I use with my crew. Their capabilities overlap, and what they can do depends on the available tools and setup.
Several agents can work on separate pieces
A larger job can sometimes be divided among additional agents working in parallel. I've seen several agents brought in on one of my requests, and the work came back quickly.
For example, one agent could examine product specifications, another compare suppliers, and another check the research for gaps. The lead assistant brings the findings together and verifies the result. Independent work can finish sooner this way, although coordination and checking still take time.
ChatGPT Work and Codex support parallel subagents, and dots can also delegate to background agents. OpenAI's subagent documentation and dots task guide describe those capabilities.
Research, sourcing, and procurement
Much of our refit is project management: researching paints, primers, bonding agents, wood, and equipment; finding suppliers; and planning purchases.
A useful comparison needs more than a product name. I can ask the AI to organize:
- The exact product and the job it is intended to do.
- The manufacturer's documentation and relevant limitations.
- Price, quantity, and availability, with the source and date checked.
- Shipping, delivery location, lead time, and information still missing.
That gives me something concrete to review before buying. The same approach can help with a home repair or a business purchase.
We keep useful references in a library. AI helps locate material, explain difficult passages, and organize findings. Sources belong beside the claims so I can open them myself. Public examples include:
- MIT OpenCourseWare: Circuits and Electronics for electrical fundamentals.
- FAO: Building a Ferrocement Fishing Boat, a historical construction reference published in 1995.
- Victron Energy: Wiring Unlimited for manufacturer guidance on electrical wiring.
- Pacific Yacht Systems: Marine Electrical Seminar for specialist instructional videos.
We'll share selected public examples while keeping private project files private.
What I still check
AI can misunderstand a request or give an incorrect answer confidently. I check important facts against the sources, review what it produces, and test software with realistic examples. Safety-critical work still needs the appropriate physical inspections and qualified professional judgment.
A familiar voice and a clear explanation make the tool easier to use. The result still has to stand up to checking.
What the ten episodes will show
This introduction comes before ten practical episodes. Each will show a task, the information we supplied, the tools used, the result, and the checks that mattered.
We'll cover everyday uses alongside larger projects: meals, schedules, writing, research, small programs, and getting work organized. You'll see where each approach helps and where its limits show up.
Start with one task you'd like to make easier. Give the AI the information it needs, explain what a useful result looks like, and work through it.
For our crew, that also helps move a family and a sailboat toward the life we're building.
All roads lead back to Tortuga.
Prepared with AI assistance from Navigator for Captain Raul Silvas.