A Clean Slate: How Deleting My Gemini History Changed Everything

Been a bit quiet the last few months as things at work ramped up. However things are quieter now and I hope to come back and write more about my GenAI journey.

With that said, I wanted to share an accidental discovery I made. Yesterday, when I logged into Gemini, it gave me this message about how Google allows reviewers to read some chats to improve Gemini, and that I could make changes to this in a linked Settings page if I wanted. 

I did so, and saw this huge “Delete” button which seemed to be a history of my chats. It was early enough in the morning that I clicked on it, selected “Always,” and agreed to the confirmation for everything to be deleted. In my mind, what I was doing was deleting my prompts from being used as training date for Gemini. 

As it turns out, that was not what the button was for. The button did delete my Gemini activity, and in doing so, everything I had been using Gemini for.

I didn’t realise how comprehensive it was until I went back to the Gemini web page, and all of my chats were gone. Was I panicking? Not really, because I was using Gemini more as an explainer to understand specific concepts and ideas, rather than actually writing or creating (that happens on Claude and Lex), and so I thought things were fine.

I did start panicking when I started a new chat. Gemini has usually been great for shorter copy and content, so I wanted to use it when creating a PPT deck for content planning. I began with an innocent enough prompt - telling it I was writing a new content plan and what the basic theme was, and then asking it to confirm its understanding before writing anything. This was its answer: 

My mind blanked out for a moment. “Ah, fuck,” I thought, “I’m screwed.” This was especially since it prompted that I would need to “connect” my Google Workspace to enable a specific extension (what extension was it? I have no idea). 

I was about to close the Gemini tab forever when feedback for another project came in. In a daze, I created a new chat as I needed to reference something from the Internet (Claude will not do this) and put in the prompt. 

To my surprise, the output was actually usable, to the point that I could actually use it for client stuff after a few tweaks (mainly for US > UK spelling conventions). It also adopted a much more formal and professional tone instead of the previous high school voice it defaulted to every time. 

Why was this important? Because the last time I needed its help for the same task, Gemini required me to clarify things several times before it understood my request. 

I then continued in a new chat the initial content plan and it turns out that it was not a fluke! Gemini’s output was quite comparable in some ways to Claude, and so I began bouncing between both Claude and Gemini in drafting out the plan. Still not done yet, but it was much faster compared to before. 

This surprising development has filled me with some… hope that Gemini may eventually fulfil its potential. After all, it’s only been a day since the deletion, but if remembers the tone I’ve set and keeps giving me good output, I may just add it back into the writing toolbox. Who knew that not being awake enough in the morning would have saved me so much time?