Over the past year, two things have been true at the same time:

• AI capabilities are improving fast
• Access cost and friction are dropping

Apps and APIs are maturing. AI will embed deeper into business software, taking on generation, recommendation, translation, summarization and other high‑frequency assistive work.

The question is no longer "use or not", but "how to use it safely and controllably".

My Daily Use Cases

Coding

ChatGPT generally understands programming tasks well and can even emit Mermaid diagrams.

• For algorithm or "toy" tasks (max array value, longest substring, basic HTTP server in Go, simple crawler in Python), it produces usable code
• With prompts and chain‑of‑thought, you can refine outputs step by step

Copilot’s in‑IDE completion pushes this further:

• When ChatGPT’s snippet is off, paste and partially comment it out; Copilot uses the surrounding context to suggest better code
• In the IDE, you can often just stub a function or write a comment and Tab your way forward
• Copilot is context‑aware at file and project level, so generated code tends to fit your style and conventions

Writing Articles / Docs

Writing also tends to follow: idea → outline → expand.

AI fits into several slots:

• Build outlines and adjust emphasis (role‑casting prompts: "You are a domain expert in finance…")
• Expand and refine paragraphs: write a rough draft, then ask AI to extend or tighten
• Run style checks, generate summaries and blurbs

Copilot helps with documentation as well:

• Boilerplate, parameter lists, code blocks in Markdown
• Docs then loop back to guide later coding work

Translation and language learning are almost trivial now.

Search

Before Perplexity, there were already RAG‑style search engines. They:

• Interpret your question better
• Pull more relevant documents from noisy result sets

Sometimes the AI answer is good enough and you never click through.

The same pattern applies to internal docs:

• Index content and run AI‑backed search
• Let AI summarize or point you at the right place instead of grep‑driven archaeology

Keeping the Boat Steady

What I thought AI would do: laundry, cooking, cleaning – the dirty, boring chores.
What AI actually does: writing, drawing, translating – the things that look a lot like my job.

Because GenAI is contextual generation, it hits word work first. For me personally, that’s coding, writing, translation, planning.

At first I was amazed at the speed and accuracy. Dependence went up; so did anxiety.

If AI keeps getting better, will anyone be left?

Two concerns are real:

1. If AI does a large share of the work, there’s less left for humans. Either the economy must grow fast enough to create new work, or some people will be displaced.
2. If we rely on AI too much, we think less. "Learning without thinking leads to confusion; thinking without learning leads to peril." Progress requires both. Remove practice and reflection, and cognitive and creative capacity erodes.

Context windows are now at 200k tokens – enough to stuff an entire classic novel in. More compute and richer models will push capabilities further.

Knowledge that took you weeks to learn might take an AI seconds.

Riding the Wave Instead of Drowning

If this makes you anxious, you’ve already taken the first step – you’re thinking about it.

Anxiety is a side effect of asking the right questions:

• How should we respond to AI’s progress?
• How do we avoid being replaced by it?
• How do we use it well?

When AI Peaks, Who Thrives?

• On one hand, AI saves a lot of time and effort
  • For me: +10–30% efficiency at work, +40% time savings outside work
  • Information gathering is much faster: define the goal, then have AI help you search and synthesize
• On the other hand, AI creates more opportunities and challenges
  • Any place where it improves efficiency has value: tech, real estate, finance, customer service…
  • It accelerates: integrating existing knowledge and learning new knowledge

"Copilot everywhere" is a reasonable prediction. The question becomes: What do you build on top of that?

A Few Grounding Points

• Today’s GenAI still has big limitations
  • Most products are text↔text or text→image; there’s a lot of sameness
  • Real commercialized multimodal hardware / robotics is rare
• Human cognition doesn’t instantly level up
  • Adoption and usage patterns differ a lot by age, region and culture
• Many industries are not fully digitized yet
  • Without clean data and processes, AI can easily amplify the wrong things
• Your own thinking still matters most
  • With family you don’t reach for AI; you talk, walk, cook
  • At real career bottlenecks, AI won’t give you a life plan

AI is an amplifier, not a replacement.

If you treat it as a leverage tool instead of a competitor, the strategy shifts from "outrun AI" to "outrun other humans with AI".