• bomewish 11 hours ago

    This exists in CeLLM and various other implementations. Well done on a smooth version. The sad thing is that when one clicks in an LLM populated cell, one cannot see/edit the content as if it were normal text. One only gets the formula. So sad. Anyone think of a workaround ? Utility would be 10x otherwise.

    • allard_eric 14 hours ago

      Amazing work! The AI integration into Google Sheets with a simple formula is a game-changer, and completing it in 24 hours is impressive. The fact that 80% of the coding was just prompting AI highlights how fast development is changing. Excited to see where AISheeter goes!

      • artur_makly 2 days ago

        thanks so much for breaking your process down like that. I've been VERY curious about this for some time.. but have not had any time to play around myself.

        In terms of the 20% traditional coding, was that something you had to figure out 100% yourself? or can a newbie also be guided on this part as well? I totally understand the benefit of just doing it quickly yourself, especially if you already have something to reuse that works.

        This is how I will be teaching my 11yr old son to code going fwd!

        re: Ai/Sheets Can it propose and generate a better "P&L" based on feeding it "myShittyAttempt" and the "Ideal_P&L_Sheet"? Would using Vision even help in this instance?

        • tuantruong 2 days ago

          Those 20% can sure be helped by AI , but since I got the coding skills few years back, it was better for me to put in the logic myself,save more time.

          Yes, in term of aisheeter, will add more items once I have more requests from users, Vision is a great idea that I could try .

          Thanks

        • miguelfernandez 2 days ago

          This is awesome! Integrating multiple LLMs into Google Sheets sounds super useful, especially for quick AI tasks. How's the performance with larger datasets, and are there any plans to add more AI models down the line?

          • tuantruong 2 days ago

            Thanks, The performance depends on the provider like OpenAI, Claude, Groq and Google (Gemini), so it shouldn't be a problem. Let's see the interests I got the users then I will plan to add more models.