Pixels > letters when it comes to LLM input. Portfolio of open-source multimodal developer tools: image-input-first creative IDEs (visual canvases for steering AI edits from reference images), reproducible image generation workflows (run receipts, evals, preference testing), and LLM observability (OpenTelemetry).

Keywords: image IDE, image-input-first, visual prompting, reference-image editing, image generation, text-to-image, multimodal dev tools, model routing, model evals, preference testing, run receipts, observability, OpenTelemetry, developer tools

• Brood

  • Desktop creative IDE for image generation (PTY-backed chat + live canvas)
  • Preview:

    

• PARAM FORGE

  • Terminal UI (TUI) for text-to-image runs with reproducible receipts
  • Preview:

    

  • Full video: PARAM FORGE preview video

• oscillo-vision-reports

  • Pairwise preference reports for vision models (PDF + LaTeX snapshots)

• superjumbo

  • Sports prediction / picks web app (React + Django)
  • Wager Town on Product Hunt: Product Hunt

• openllmetry (fork)

  • OpenTelemetry instrumentation for LLM providers (Traceloop fork)

• LinkedIn: https://www.linkedin.com/in/kshowkat/
• Email: showkatkevin@gmail.com