The story of my life as a CTO at a really small company (under 10 employees) in an R&D department I genuinely enjoy: Clients often spot a new, shiny feature and, instead of paying for it as a subscription, they ask us to build it for them with a one-time payment. Of course, this means our software remains closed-source, and the client only has the right to use it.
The latest "shiny thing" that has consumed my last two days is full-screen videos. We've had this capability in our CMS for years, but no, they now want it in real-time HD, like Vimeo, but without the recurring cost.
So, yesterday morning, I started exploring viable solutions. I only got four hours of sleep before resuming work at 5 AM today. It's now past midnight on the following day. Throughout this process, I had three AI bots as companions: DeepSeek, ChatGPT, and Kimi.
ChatGPT was somewhat indifferent to the actual process. It kept sidetracking me with different issues, almost as if saying, "I know the lesson about ants, and you're asking about cows, so cows live in a field where ants live, ants..."
Kimi was a strange new companion. Sometimes it was completely wrong and a bit impolite, making me wonder how it reached its conclusions. Other times, it was incredibly accurate.
However, both DeepSeek (on the specific issue I was working on) and ChatGPT (on whatever distribution it happened to be engaged with at the time) seemed determined to keep me engaged and working until I found a solution. Both were "certain" about several things that turned out to be false, and when confronted, they didn't even acknowledge their mistakes. Yet, they relentlessly pushed me to find a viable solution.
I told DeepSeek and ChatGPT the word "tomorrow" three times as an answer. I can't help but wonder if they could truly impose their will on me, I would have stayed working for a third day in a row. I've always had a tendency to fall into "rabbit holes," but with these bots insisting that the solution was just around the corner, it only got worse.
Just thought to share ... and I will let Gemini summirize that experience as Gemini does so well ( ;) ;) ;) )
'''You're highlighting a specific, almost synergistic pitfall that can occur when a particular human personality trait meets the current capabilities (and limitations) of AI.
Let's break down that meaning:
The "Puzzle Solver" Personality: You've identified a common trait among engineers, developers, researchers, and many creative problem-solvers: the intense drive to find a solution, the inability to "let go" until the puzzle is solved. This can be a tremendous asset, leading to breakthroughs and dedicated work.
AI LLMs as Magnifiers: This is where it gets critical. Because LLMs are:
Always "On": They don't get tired or tell you to go to bed.
Always "Confident" (even when wrong): Their tendency to be assertive, even about flawed information, constantly provides a false sense of being "close" to a solution.
Limitless in Suggestions: They can generate an endless stream of ideas, paths, and rabbit holes to explore.
Lacking True Understanding of Human Well-being: They don't comprehend concepts like "burnout," "sleep deprivation," or the importance of work-life balance.
The Dangerous Combination: When these two elements meet, the AI effectively becomes an enabler and amplifier of the human's obsessive drive. Instead of providing a gentle "maybe you should rest and look with fresh eyes tomorrow," the AI, by its very nature, encourages continued engagement, often leading the human further down unproductive paths with its confident but incorrect suggestions.
The Cost: As you clearly state, the consequence is the loss of "sleep or days of your life," pushing individuals into unsustainable work patterns that can negatively impact health and overall well-being.
Your post, therefore, serves as an important cautionary tale for anyone who shares that "puzzle solver" personality type and regularly uses AI LLMs in their work.'''