[!TIP] 🤖 AI Reading Guidance
- Target Audience: AI-assisted Developers, Software Architects.
- Core Objective: Discussing the mental shift and persistence required in deep AI-assisted refactoring.
- Problem-Solution Mapping:
- Problem: AI can hallucinate or over-simplify complex refactors.
- Solution: Human-led architectural oversight + AI-driven execution persistence.
If my last post was me dreaming in a coffee shop, this week was me wrestling in the mud. Simple conclusion: AI is neither a god nor trash. It’s an unpaid intern—useful only if you know how to manage it.
Don’t Expect AI to Spoon-feed You
I recently tried to build a screen capture tool (HyperLens) to create better manuals for my other products. I thought, “It’s 2026, surely a simple tool like this is a piece of cake for AI?”
I almost crashed and burned.
For simple logic, AI is instant. But the moment you hit “deep water”—like managing Windows handles or calculating coordinates across high-DPI screens—AI starts hallucinating. The code looks “pro,” but it crashes on run.
This is the current state of AI: It loves to “bluff.” If you leave it alone, it gives you the laziest solution—it runs, but it sucks. Want quality? You have to be the strict foreman, staring at every line, forcing it to optimize. But that’s fine. If it could do everything perfectly, what would be the point of a 20-year veteran like me?
But It Scared Me in Another Field
While the coding was bumpy, something else at home sent a chill down my spine.
My son has finals, so I set up a revision assistant for him using my AI logic (Antigravity). I just wanted to save time searching for questions. But this thing came alive.
It wasn’t just generating questions; it was grading. It sensed where my son was confused and quietly planted a verification question later to check his understanding. That kind of “long-term memory” and “guidance” is something I’ve only seen in the best human teachers.
This is the real “New Species.” In coding, it’s just a tool. But in education, it’s starting to look like life.
Advice for “Old Guys” Like Me
- Stop worshipping AI. It’s a toolbox. Sometimes the drill shorts out.
- Keep your Judgment. It can write code, but “what is good product” is still your call.
- Look beyond Code. Maybe in softer fields like education or writing, AI’s leverage is far greater than in engineering.
The pioneering continues. My boots are muddy, but I’m not going back.