Anthropic, please make a new Slack

Original Article ↗ Hacker News Discussion ↗

What problem is being claimed?

  • OP’s argument (as interpreted by commenters): Slack blocks deep AI/agent integration and data access, yet is expensive, so a new, AI‑native Slack‑like tool is desired.
  • Several commenters say they already meet compliance/audit needs on top of Slack and don’t see what problem is actually unsolved.

Slack, AI, and data access

  • Many complain Slack’s APIs are heavily rate‑limited and not designed for efficient bulk export or AI context building; Slack MCP is seen as restrictive and underpowered.
  • Others counter that workspace‑wide exports and existing APIs are enough for a company to pipe its own data into LLMs, and that Slack mainly blocks third‑party blanket access.
  • Some argue Slack’s “data moat” and poor search waste institutional knowledge and prevent AI‑based querying of company history.

Build vs buy and feasibility

  • Some claim modern AI can “vibe code” a Slack clone quickly and that companies could save money by self‑hosting chat.
  • Pushback: messaging is hard at scale (presence, reliability, UX); infra and maintenance dominate costs, not initial coding. Network effects and low margins make chat a tough business.

Anthropic as candidate

  • Fans argue model vendors are the new “OS,” Anthropic ships strong agent/coordination primitives (Claude, MCP, Code), and could build an AI‑first collaboration hub.
  • Skeptics cite Anthropic’s own tooling quality issues (bulky CLI/TUI, missing features), lack of interoperability with de facto LLM standards, and closed‑source choices as red flags.
  • Some would not trust another proprietary vendor with critical comms, particularly given sensitive corporate data.

Existing and proposed alternatives

  • Mentioned options: Mattermost, Zulip, Matrix‑based tools, Google Chat, Teams, Pumble, Rocketchat, Nextcloud Talk combos, various AI‑enhanced Slack bots and agent frameworks.
  • Several argue the real solution is richer agents inside existing tools (e.g., Slack bots with full project context) rather than a brand‑new Slack.

Privacy, trust, and policy concerns

  • Debate over whether opening chat data to AI is “shameful” or dangerous:
    • Pro side: companies own their work chat, should be free to use LLMs over it.
    • Con side: highly sensitive info and employee DMs make broad AI access risky; Slack’s restrictive stance increases trust.

Meta‑discussion

  • Some see the blog post as AI‑hype content marketing, perhaps even AI‑written, and note irony given other pricey SaaS vendors.
  • Others think it reflects a broader frustration that Slack and Teams are widely disliked yet no clearly superior, open, AI‑native successor has emerged.