Delta
Zed’s new Delta tool aims to make “multiplayer AI” a core part of software development, combining shared agent chats, inline annotations, and a DeltaDB-backed transcript of how code changes were produced. Commenters see potential for mentoring, debugging, and coordinating agent-driven work across teams and models, especially through richer ways to comment on and navigate long AI-generated plans. At the same time, many are skeptical about real-time collaborative editing, long-term lock‑in and privacy implications of storing every agent conversation, and Zed’s apparent shift of focus away from polishing its core code editor toward increasingly AI‑centric products.
AI Explanations and Code Summaries
- Many dislike default LLM code summaries: too verbose, miss edge cases, or feel like “intermediate dev” prose.
- Some find them useful when tightly scoped: high-level overviews of diffs, schemas, APIs, with humans filtering and polishing.
- Several advocate better formats than long prose: pseudo‑code, “toy examples,” structured spec languages, walkthroughs with highlighted snippets, or even comics.
- A recurring pain point is having to iterate prompts multiple times to get a concise, accurate explanation.
Delta / DeltaDB Concept and Value
- Core ideas perceived: multiplayer agent conversations, inline annotations on agent output, and conversation-as-document.
- Supporters see strong use cases for mentoring, handoffs, debugging agent output, and cross-model workflows.
- Skeptics question differentiation vs existing tools (Codex, Cursor, CLIs, other harnesses) and view it as another “agent window” that can be cloned.
- Some question DeltaDB and transcript storage as overkill compared to Git workflows or code-centric context.
Saving Agent Conversations vs Documentation
- One camp sees persistent transcripts as crucial: enable resuming work, recovering intent, verifying completion, and querying past reasoning.
- Others argue long chats are meandering; what matters is distilled specs/ADRs. They fear this will replace rather than improve documentation.
- Some worry that LLM-generated docs become verbose timelines instead of clean, updated specs.
Multiplayer Coding and Collaboration
- Many commenters are skeptical of real-time “multiplayer” coding; pair programming and live-sharing are seen as niche or exhausting.
- Others report positive experiences when pairing works well, especially for teaching, tricky problems, or debugging.
- Some think Delta’s real value is asynchronous shared context with agents, not simultaneous editing.
Zed’s Direction and Product Concerns
- Mixed sentiment: praise for Zed’s speed and innovation in human–AI interaction, but frustration that core editor bugs and basics feel neglected.
- Some see an “identity crisis” or AI pivot driven by VC pressure, with fears of lock‑in and potential future license changes.
- A subset prefers non‑AI or OSS alternatives (Gram, neovim, Sublime, etc.) and resents AI becoming central.
UX and Accessibility Feedback
- Multiple complaints about low-contrast design, hard-to-read themes, and vague product descriptions without clear demos.