IBM Bob

Original Article ↗ Hacker News Discussion ↗

IBM’s new “Bob” tool, a VS Code–style AI coding assistant that routes requests to various underlying models and bills usage in proprietary “Bobcoins,” is being met with widespread skepticism. Commenters question its unclear value compared to established tools like GitHub Copilot and Claude Code, criticize opaque model selection and pricing, and see it as another late, branding-misguided attempt by IBM to stay relevant in AI. Some note potential niche utility for IBM mainframe and legacy Java stacks, but doubt it will gain traction or materially change IBM’s long-term trajectory.

Overall sentiment

  • Predominantly negative or mocking reaction to IBM Bob’s launch.
  • Many see it as late to the AI assistant party and emblematic of IBM’s decline in product quality.
  • A few self-identified IBM employees or users say Bob is helpful and integrates well with IBM tooling, but others who tried it call it “dreadful” or a “dumpster fire.”

Product nature and capabilities

  • Widely perceived as a VS Code fork with an “agent” side panel.
  • Several commenters say it’s mainly a harness that routes requests to various LLMs, possibly including IBM’s Granite models, with no user control over which model is used.
  • Strength is assumed to be in IBM-specific domains (mainframe, IBM i, Java/WebSphere, legacy modernization) via paid “skills” or workflows.
  • Unclear from the discussion exactly which models run where and how much is proprietary vs hosted open weights.

Branding, naming, and design

  • Heavy criticism of the name “Bob,” due to:
    • Association with Microsoft Bob (seen as a historic failure).
    • Childlike / “Bob the Builder” vibes and UK slang (“bobbins” = rubbish).
  • Landing page and mascot are seen as off-brand and poorly designed compared with IBM’s own Carbon design system and Plex typography.
  • Testimonials read like AI-generated or corporate boilerplate and are mostly from executives, not working developers.

Pricing and “Bobcoins”

  • Billing is via “Bobcoins,” a proprietary consumption unit.
  • Strong dislike of game-style virtual currency: seen as obfuscating real costs despite claims of “transparent, predictable” pricing.
  • Confusion about what exactly a Bobcoin buys and whether it can be “earned” or only purchased.

IBM’s strategy and reputation

  • Many view IBM as a services/consulting-heavy company clinging to relevance by rebranding others’ models and adding markup.
  • Skepticism that IBM will maintain Bob long-term, citing abandoned or sidelined tools (Watson, RDi, SEU).
  • Some push back, noting long-lived IBM platforms (mainframes, MQ, Db2) and niche strengths (legacy compatibility, enterprise LLMs like Granite).

AI ecosystem and competition

  • Compared to GitHub Copilot, Claude Code, Cursor, and others; Bob is seen as lagging and undifferentiated.
  • Some argue harnesses are commoditized and that competition may still be beneficial.
  • Question whether enterprises will prefer IBM-branded tooling for governance/cost-control versus best-of-breed first-party AI tools.