Roadmap

Introduction

In my pursuit to develop a prototype for a seedful and seamful search interface, this site is continuously evolving and under construction.

Current State of the Site

• View previous updates in the change notes.
• For more details, visit Site and search/About.

Roadmap Overview

Content

• Write: [Re Berjon's “Fixing Search”]: Fixing Search

Code

Adding Additional Search Backends

• Transition Lunr.js to a working example and introduce new search backends:
  • Whoosh
    • Note: I’ve been using Whoosh for my personal note searches.
  • Algolia
  • Elasticsearch
• Enhance external search capabilities.
  • Current status: Basic. See search/About or search for type:external.

Retrieval Augmented Generation (RAG) Search

• Implement search with retrieval augmented generation (RAG).
  • Progress: Basic offline example available: Asking “Why hire Daniel?” in an LLM search w/ local Lunr.js.

Transparency, Feedback, & Curation

• Enhance feedback mechanisms.
  • Progress: Basic feedback mechanism available: Feedback.
• Implement curation workflow for the dynamically generated snippets.
• Publicize the GitHub repository for this website.

Accessibility Improvements

• Testing with Lighthouse and Other Audits

Projects

SearchRights.org

• Progress:
  • initial write-up: Planning CSUIRx
  • website: (under construction) SearchRights.org
  • GitHub setup (github.com/danielsgriffin/SearchRights/)
• Next Steps:
  • Add search via danielsgriffin.com

Loosely-functional speculative designs in userscripts:

ctplsm (Contextual Twitter Poultice for Learning So Much)

• Progress: Functional prototype, on GitHub.
  • github.com/danielsgriffin/ctplsm/
  • Added prompt-and-response logging.
• Next Steps:
  • Add README
  • Log user feedback
  • Modularize ‘Check elsewhere’

Setarco

• BLURB:

  This is a userscript that randomly inserts (or interjects) an additional instruction in a prompt on ChatGPT conversations (matching https://chat.openai.com/*). Setarco lets users set engagement to automate randomizing custom orders. The additional instructions are written to encourage questions from that LLM-system that are generative, though provoking, revealing, surprising, and perhaps even challenging.

• Progress: Functional prototype, on GitHub.

  • github.com/danielsgriffin/setarco/

• Next Steps:

  • Log prompts-and-response and user feedback

Auto-Button-Click Double-check Evaluation for Funsies or Gnoses

• BLURB:

  I want to write a userscript that automatically clicks the “G” Google Search button. So perhaps the evaluation could appear while I’m glancing at the response. (It does take a few seconds?) [on Twitter]

• Progress:

  • initial experimentation…

Tool Exploration

• GitHub Copilot Chat
• Sourcegraph
• Cursor
• ChainForge
  • Progress: A couple attempts and a post: The Need for ChainForge-like Tools in Evaluating Generative Web Search Platforms