Mindset
- Future
- Chatting with VSCode core developer Lu Peng about AI programming, Vibe Coding, and whether programmers will still exist in the future
- Writing programs is like riding a horse; we are the last generation of horse riders.
- DHH: Future of Programming, AI, Ruby on Rails, Productivity & Parenting
- no one knows anything, and you have to then position yourself for the future in such a way that it doesn’t matter. you pick a professional path where if it turns out that you have to retool and re-skill, you’re not gonna regret the path you took.
- There will be more one-off code in the future.
- Chatting with VSCode core developer Lu Peng about AI programming, Vibe Coding, and whether programmers will still exist in the future
- Vibe Coding
- Vibe-coding is not the same as AI-Assisted engineering.
- Using “Vibe coding” to build a real product won’t work. But it’s great for prototyping.
- AI won’t kill software engineering; instead, it makes the old rules of software engineering great again.
- In the era of AI-assisted programming, automated software testing is becoming increasingly important.
- Vibe Coding Is Creating Braindead Coders
Tech
Spec-Driven Development
Advice
- Study with AI
- GenAI policy in Berkeley CS 189 2025 Fall & jyy OS 2025 Spring
- For repetitive or formulaic code, like data visualization, you can directly tell the large model what kind of chart you want and let it generate it for you. But for the core model implementation and algorithm logic, it’s best to write it yourself to greatly enhance your understanding. Source
- (2025.9.19) Use GPT-5-high for more macro-level planning, and GPT-5-Codex for specific technical implementation.
- Use
git worktreefor parallel tasks. Related command-line tool xlaude and its workflow sharing.
Prompt
(2025.7)【Official Anthropic Share】Prompt Engineering Techniques
Example of calling Gemini in Claude Code:
- Q: “Please have gemini-2.5-pro comment on each section, mainly to add content.” How is this implemented? Calling other models in the code? A: Example given:
gemini -p "In-depth answer: <question>" -m gemini-2.5-pro. CC can understand what “Please have gemini-2.5-pro" means.
- Q: “Please have gemini-2.5-pro comment on each section, mainly to add content.” How is this implemented? Calling other models in the code? A: Example given:
Feed
Status
- 2025.9
- At the 2025 ICPC World Finals, OpenAI’s general-purpose reasoning system achieved an unprecedented score of 12/12, defeating all human teams. Source
- 11 problems were solved on the first attempt, and the final, most difficult problem was solved using GPT-5 + an experimental reasoning model.
- DeepMind’s Gemini 2.5 Deep Think model also performed well, solving 10 out of 12 problems.
- “70%, 80%, or even 90% of Anthropic’s code is written by Claude. People think this is false because they think it means firing 70%, 80%, or 90% of software engineers, but what’s really happening is that humans are becoming managers of AI systems.” Source
- At the 2025 ICPC World Finals, OpenAI’s general-purpose reasoning system achieved an unprecedented score of 12/12, defeating all human teams. Source