I went down another rabbit hole this morning, this time reading the 2026 roundups on Juejin, and the pattern that jumped out is that the AI coding tool conversation has moved on from "which one wins" to something more interesting. Everyone is still arguing about Cursor versus Claude Code, but the people shipping real work have already moved past that question.
The thing I keep noticing is that the admired-vote gap from the Stack Overflow 2025 survey is doing a lot of work. Claude Code at 46 percent admired, Cursor at 19, GitHub Copilot at 9. I want to take that with a small grain of salt because the survey population is self-selecting, but the direction matches what I'm hearing from people I actually trust. Cursor is the daily driver for most folks I know who pay for a tool. The Tab flow, the inline diff UI, the multi-file edits in the agent tab — it's just faster for the boring 80 percent of the work. Claude Code is what they reach for when the change spans more than they want to think about, especially the legacy stuff with weird business logic baked in over years. The terminal-first flow with planning, sub-agents, and verifiable steps is genuinely different from anything else, and the people who do big refactors keep telling me it has changed how they work.
What surprised me reading the Chinese lists this time is how much the domestic models have closed the gap. 通义灵码 from Alibaba is showing up everywhere with a pitch that it matches GPT-4o quality at lower enterprise cost, and the integration story for companies that already live inside Alibaba Cloud is brutal to beat. DeepSeek-V3 keeps coming up as the model underneath a lot of newer tools, and the cost math is hard to argue with if you're running inference at scale. I have not stress-tested any of these in production myself, so I'd take my opinion with a grain of salt, but the rate of improvement from 2025 to now has been faster than I expected.
The post that actually stuck with me this morning was a Juejin piece arguing that the next bottleneck is not the model or the IDE — it's the input. The author made the case that as Cursor, Claude Code, Copilot, and the Chinese tools converge on roughly the same capability set, the real productivity gap shows up in how you structure what you hand them. Mermaid diagrams for data flow, explicit API contracts, generated ER diagrams, a written spec for the change you're about to make. Documentation-driven development as a way to feed the agent, not as a relic of waterfall. Honestly I had been feeling this for months without a name for it, and reading the article felt like someone describing the thing I was already doing badly and finally making it explicit.
To be fair, I'm not fully sold. There is something performative about writing a spec for a two-file change, and most of my day is two-file changes. I think the right version of the argument is that for anything that touches more than three files or lasts longer than a week, the upfront spec pays for itself, and the tools that handle that input well are the ones I'll keep using. The agent will keep getting smarter, but it cannot read your mind about which tradeoffs you care about, and that's the part that won't be automated away.
I'll reassess in three months again. That's the only honest timeline I can commit to, and the last time I said that I was using Cursor as my primary and bouncing into Claude Code for big refactors, which is still roughly where I land today. The tools are better, the ecosystem around them is bigger, the domestic options are real, and the workflow is what actually differentiates you.
