823: Is Cursor AI the VS Code Killer?

Episode Summary

Executive Summary: The episode evaluates Cursor, an AI-enhanced code editor built on VS Code, and concludes it is the best AI coding assistant the hosts have used so far—especially for UI, context awareness, and multi-file edits—while also cautioning that it is overhyped and not a replacement for a developer. They compare it to GitHub Copilot, discuss pricing, model selection, docs indexing, and its limits, then end with unrelated “sick pics.”

Main Topics: Cursor vs. GitHub Copilot: The hosts compare Cursor favorably to Copilot, saying it feels significantly better in real use, especially for context-aware assistance and a smoother workflow. UI as the differentiator: A major point is that Cursor’s custom interface layer—smart rewrites, richer diffs, better chat, and clearer completions—makes the experience feel more intuitive than a plugin alone. Context awareness and multi-file editing: Cursor can reference files, symbols, docs, and entire codebases to make edits across multiple files and even create new files, which the hosts see as a major step up. Strengths and limitations of AI coding tools: The hosts emphasize that Cursor is best as an assistant/booster rather than an autopilot. It can fail badly when asked to generate whole apps, but is useful for incremental coding help. Docs indexing for newer libraries: They highlight Cursor’s ability to ingest custom documentation sources so it can answer questions and generate code for newer or fast-changing ecosystems like Svelte 5, Godot, and TanStack Router. Pricing and business value: Cursor’s $20/month price is discussed as relatively expensive compared with Copilot and Codium, but potentially justified if it saves enough time or is paid by an employer. Sick picks and closing banter: The episode ends with recommendations for a bodybuilding documentary YouTube channel and an OXO salt/pepper grinder set, plus some playful off-topic conversation.

Key Arguments: Cursor feels meaningfully better than GitHub Copilot in day-to-day use, even though both rely on similar underlying model capabilities. The biggest advantage of Cursor is not a new model but the interface and workflow design around AI assistance. It is especially strong when it can see the codebase context, reference symbols, and apply edits across files. Cursor is useful as a second brain or productivity booster, not as a tool that replaces the developer. Indexing documentation makes AI coding assistance much more reliable for new or fast-moving libraries. The product is probably overhyped online, but not garbage; the hosts consider it the biggest leap in AI coding UX in about a year. The $20/month price is high for individuals but more plausible as a business expense or team tool. Attempts to have Cursor build a whole React app from scratch exposed its limits and showed that human oversight remains necessary.

Data Points: Cursor monthly price: $20/month - Discussed as the Pro plan cost for unlimited completions and 100 fast premium requests. Fast premium requests: 100 - Included in the Pro plan before requests may be queued at high load. GitHub Copilot price: $10/month - Used as a lower-cost comparison point to Cursor. Codium price: $15/month - Mentioned as another comparator in the coding assistant market. CodeWhisperer price: about $15/month - Estimated competitor pricing discussed alongside other assistants. Anthropic context window: 100,000K context window - Referenced while explaining why larger context windows improve coding assistants. Cursor funding: $60 million - Mentioned as evidence the company has serious backing. Bodybuilding videos on cited channel: about five videos - The sick-pick YouTube channel was noted as small but high quality. Bodybuilding documentary length: 35 minutes to an hour - The Iron Historian videos were described as mini-documentaries of this length.

Pivotal Quotes: "I would say, significantly better than GitHub Copilot, which is saying something." — Wes Boss: Wes explains why he is impressed with Cursor compared with the tool he already uses. "The value in this is how they are scanning your code base and architecting and to me quite honestly the value here is the UI that they're doing." — Scott Tolinski: Scott argues Cursor’s interface and codebase-aware workflow are the real differentiators. "It's a booster pack and it will help you be a better developer." — Scott Tolinski: The hosts frame Cursor as an augmenting tool rather than a replacement for developers.

Implications: Cursor shows that AI coding tools are now competing on UX, context handling, and workflow integration, not just model quality. Expect VS Code/Copilot to copy these ideas quickly, while developers will still need skepticism and review.