Fooled By AI

If AI is lying to you, how do you know?

Think about that. If you don't know anything AI is "talking about", basically you are in deep trouble. You must be able to validate, fact check and even disagree with AI output. High dependency on AI is bad, we should be able to program if we don't have internet or are running out of tokens.

Do you know how to know when a human using AI is fooling you? Do you know if you are fooling yourself with AI?

Take a look at the Clever Hans effect.

The dark side of AI

AI can be and is used for a lot of scams. AI and people using AI can fool you in many ways. AI has a dark side. Perhaps many dark sides.

Outsourcing Gym Workout

Think about this: we could easily pay someone to go to the gym for us. That person can do push ups, run, lift weights. Be at the gym everyday and at the same time we never be at the gym. Genius! Well, the issue is that your "contractor" will be strong and in shape but you will not. Although you can claim the glory of going to the gym, you do not have the benefits, you only have status.

With AI it's the same thing. Anyone can use AI to generate code, does that mean we are done there? Think about it, everybody has the same AI, same models, same capabilities. So just using AI is not differentiating.

Solutions vs Wrappers

Now, many startups are building or vibing something with.ai or a new Agentic. Under the hood most of the time it's just a wrapper around OpenAI API or Anthropic API. So what is the real value add?

I love Rappers but I never liked Wrappers. In my book Principles of software architecture modernization I wrote about the dangers of wrappers with internal shared libraries. Now we have the same danger but with external tools. There is an explosion of tools, new tools, all old tools being rebranded to AI, or AI features popping up in software we would not expect.