Gemma 4 Multimodal Fine-Tuner vs IsItAgentReady

Gemma 4 Multimodal Fine-Tuner is an open-source toolkit that lets developers fine-tune Google's Gemma 4 and 3n models across all three modalities — text, images, and audio — using only Apple Silicon hardware. It runs natively on PyTorch with Metal Performance Shaders (MPS), bypassing the NVIDIA requirement that has historically blocked Mac users from serious local fine-tuning work. The toolkit handles the full training pipeline including dataset prep, LoRA adapters, and multi-modal data collation. It ships with working example notebooks, a validation suite, and clean abstractions that don't require deep familiarity with the underlying MPS stack. Apple Silicon's unified memory architecture actually helps here — large multimodal batches fit in memory that would otherwise require GPU VRAM splitting on CUDA setups. Posted to Hacker News on April 7 as a Show HN, it pulled 109 upvotes and 165 GitHub stars within hours. The timing is sharp: Gemma 4 just dropped days ago with new multimodal capabilities, and the community immediately wanted local fine-tuning. This fills that gap faster than Google's own tooling.

IsItAgentReady is a free web scanner that audits any URL for AI agent readiness across 36 checkpoints organized in five categories: robots.txt compliance (covering all 13 major AI crawler bots), structured data (17 Schema.org types), llms.txt implementation, MCP endpoint detection, and OAuth/agentic commerce readiness. Each category gets a letter grade with specific, actionable fix instructions. The tool was built by a two-person team responding to a growing pain point: as AI agents replace search engine crawlers as the primary way content is discovered and consumed, most websites are not configured to be agent-accessible. A site might have perfect SEO but actively block Claude, GPT, or Perplexity crawlers in its robots.txt — effectively invisible to the AI-driven web. IsItAgentReady surfaces these gaps in about 15 seconds. It also ships as an MCP server, making it usable directly from Claude Code, Cursor, Copilot, or any MCP-compatible environment: run a scan from the terminal and get structured results without leaving your editor. The project is positioned as "Google PageSpeed Insights for the agentic web" — a framing that resonated on Hacker News where it appeared as a Show HN with strong engagement.