本文由 AI 分析生成
Summary
A data scientist’s practical account of integrating GenAI tools into daily work across four use cases: learning, coding, documentation, and writing. Tools covered: ChatGPT, Claude, Perplexity AI, GitHub Copilot. Two years post-ChatGPT reflection on what’s changed and what the limitations are.
一位資料科學家關於將 GenAI 工具整合到日常工作的實踐分享，涵蓋四個使用場景：學習、編碼、文件和寫作。工具包括 ChatGPT、Claude、Perplexity AI、GitHub Copilot。ChatGPT 兩週年後對變化和局限性的反思。
Key Points
- Learning: GenAI as first-pass explainer; Perplexity AI for sourced answers (partly replaces Google); then dive into papers/docs old-school for details
- Coding: GitHub Copilot autocomplete for known implementations; ChatGPT/Claude for “I know the output, not the approach” problems; major time savings on boilerplate
- Shift in work focus: GenAI handles easy/repetitive coding → more time for strategic thinking and complex problem-solving
- Multiple tools needed: ChatGPT and Claude give different suggestions; hard to predict which is better for a specific problem
- Warning: quality strongly depends on prompt quality; author admits room for improvement
Insights
The two-model strategy (ChatGPT + Claude for different coding problems) reflects a real observed phenomenon: the models have different “personalities” in problem-solving and different training emphases. Using both and comparing is more reliable than assuming either is universally better. The pattern of GenAI for first-pass → old-school sources for depth is a mature workflow: it avoids both the trap of trusting AI for authoritative details and the inefficiency of starting cold from primary sources.
Connections
Raw Excerpt
I am more productive as I can iterate faster and try more things as I do not need to write all the code myself. As GenAI tools take over some of the easy but time-consuming parts of my work, I have more time to spend on strategic thinking, complex problem-solving, and being creative.