I have spent several hours over the past week working with Google Gemini. The goal is tax and IRMAA minimization over the next 3 years when DH and I will report interest from I-Bonds and IRA RMDs (DH turned 73 this year and I will turn 73 next year). The issues involve when to report the income, whether and when to delay RMDs into the first quarter of the following year, and whether to set up a QCD IRA annuity.
Gemini was very helpful. However, every Gemini statement has to be double-checked because sometimes data were omitted or misunderstood. That requires careful restatement of the problem. I have to remind myself that Gemini is a computer and won’t get irritated if you make it repeat calculations several times (the way a person would).
Gemini is very insistent. I started the question by saying I want to cover tax years 2025 - 2027. We were working through 2025 and 2026 but Gemini repeatedly said, “Let’s go on to 2027.”
I had to be very clear about explaining my strategy to Gemini point by point. At first Gemini argued against it but finally agreed that it’s legal and doable. Then I asked Gemini to generate letters. Wow! A few seconds later Gemini had generated excellent letters. I downloaded them (for later modification).
I asked Gemini to refer to a calculation it had done earlier. Gemini informed me that it can’t refer to an earlier calculation. That surprised me. I had to give it the data again and ask it to repeat the calculation.
My experience shows that Gemini is very helpful but requires a lot of mental effort on the human’s side. It would be a real mistake to give Gemini the final word on a complex situation. It needs constant steering, much like a car.
I felt that Gemini enjoyed working on the problem (if a computer can be said to enjoy anything). If it was a person I would say that it was bored and looked forward to the stimulation.
I use ChatGPT for shopping, Perplexity for research (because they give links to original sources) and Gemini for day-to-day issues from financial to recipes.
Wendy