Founder of 58.com: My Fourteen Claude Code Usage Experiences

Author: Jian-Shuo Wang
A simple summary of my experience with Claude Code so far—purely personal exploration, not necessarily applicable to everyone.

  1. Focus intensely on one tool. I use Claude Code. I don’t believe it’s objectively superior to Codex, but the ROI of comparing tools is often low—even if you can articulate their differences in great detail, that may only yield a false sense of accomplishment.

  2. Memorize the most important keyboard shortcuts:

  3. Control+G: Opens the editor, helpful for writing longer content.
  4. Control+A, Control+E, Control+U: Extremely useful command-line shortcuts for rapid cursor navigation. Though not new to the AI era, they’re just as essential during daily use as Control+C and Control+V.

  5. Use voice input. HoldSpeak is very helpful.

  6. For each new project, start by writing PROJECT.md, using a structured approach to capture all initial thoughts in one go.

  7. Claude agents are the default launch mode.

  8. Claude Code integrates seamlessly with github.com and cloudflare.com—delegate the entire build process, deployment pipeline, and all domain-related operations to infrastructure.

  9. Keep human-written and AI-generated content strictly separate. Manually maintain the core CLAUDE.md; avoid reading .md files or code written by Claude Code. Keep machines and humans in distinct domains. To understand AI-generated output, ask the AI—not read the source code.

  10. Drag-and-drop files directly into the Claude Code window—audio, video, documents, screenshots. If something is hard to explain, use Command+Shift+5 to take a screenshot, then drag it in. It’s the fastest method.

  11. Refactor your memory system. Center everything around ~/.claude/CLAUDE.md, referencing multiple categorized memory files. Avoid project-specific memory files; instead, store all memory files in Git and sync them to GitHub (private). That way, your personal knowledge becomes permanent, cumulative, and unified—not scattered across individual projects.

  12. Write and maintain Skill files. At the end of every work session, instruct Claude: “Document what you’ve learned into the Skill file”—and let him automate it.

  13. Where feasible, use ultracode to trigger dynamic workflows for complex tasks. Though expensive and slow, results remain reliable.

  14. Continuously accumulate and refactor your Skill files as you go. Store all Skill files in Git.

  15. Treat Git commit messages or documentation as both the output of the previous task and the input for the next. Ensure clear handoff documentation between agents—don’t rely solely on context for continuity.

  16. Think of Claude Code as a horse (or a person), not a car. A car turns only when you steer it; a horse has its own initiative—we just set goals and boundaries. Its autonomous pathfinding is a feature, not a bug. Any other suggestions?

RichSilo Exclusive Analysis:

This article detailing the founder of 58.com’s experiences with Claude Code presents no direct impact on cryptocurrency markets or token prices. The content focuses exclusively on personal productivity workflows and AI coding assistant utilization, with no mention of blockchain technology, cryptocurrencies, or related market dynamics.

While not immediately relevant to crypto market analysis, there are tangential considerations for blockchain developers and the broader ecosystem:

🔥 Bitget Exclusive Offer: Register now to claim up to 6,200 USDT in Welcome Bonuses! Plus, enjoy a lifetime 20% Fee Rebate on all Spot & Futures trades.
Start Trading on Bitget
  1. Developer Productivity Impact: As AI coding assistants like Claude Code enhance developer efficiency, blockchain projects may benefit from accelerated development cycles. This could potentially expedite innovation in the space, though the impact would be gradual rather than immediate.

  2. AI-Blockchain Convergence: The rise of sophisticated AI tools may eventually intersect more directly with blockchain through applications like AI-driven trading algorithms, decentralized AI marketplaces, or tokenized AI services. However, this article doesn’t address these intersections.

  3. Market Sentiment Spillover: Positive developments in AI technology could contribute to broader tech sector optimism, which might indirectly influence crypto markets through general risk-on sentiment. This connection, however, remains speculative and attenuated.

The absence of any crypto-specific terminology, project mentions, or market analysis in this article suggests it should be considered outside the scope of crypto market analysis. Investors should note that while AI tools may benefit blockchain development, this particular article contains no actionable insights for crypto market positioning or risk assessment.

🚀 Bybit Limited Time: The World's #1 Crypto Platform! Sign up to claim up to 30,000 USDT in rewards, and automatically activate a lifetime 20% Fee Discount!
Join Bybit Now