CARDS: Crowdsourced Annotated Research for Debating Solutions
Overview
Critical times call for critical thinkers to create a crowdsourced annotated research dataset, for AI models to recommend research quotes, to evolve crowdsourced chain-of-thought reasoning, unlock faster ways to read long articles, to monitor developments in a knowledge graph by topic modeling, and to provide a public service of answers to research. Debate should be a war of warrants where victories are vectorized as weights — weights which lead to the emergence of Collective Consciousness.
Features & Capabilities
Search within summaries, cites, highlighted and full text over millions of research quotes. Scout what evidence selected teams and schools frequently read. Sort by Speeches Read In Count (statistics from over a decade in all styles). Use Words Bold & Highlighted to match speech times (~300 w/min) and to read aloud warranted summaries while judge & opponents reads highlighted quotes on-screen, which makes context more understandable and publically persuasive. Collaborate on topic outlines, share evidence, and download Topic Starter outlines.
Recent quotes, outlines and debates
The Quotes, Outlines and Debates buttons under the search box choose what the list shows. Quotes are listed most recently popular first — the current season's cards, the most read within it, newest card first on a tie; the Outlines and Debates buttons list the most recently uploaded documents and round videos, and typing a term narrows any of those lists. With a term, quotes are listed best match first instead. The search opens on Quotes with the Recent & Popular sort, so cards are there without searching. Sorting by Most Read or Season overrides the recent-popular order for quotes.
Search tips
Results are ranked by relevance: a card whose tag names your words comes before one that only mentions them in its body, and a card containing more of your words comes before one containing fewer. Put words in double quotes to match an exact phrase, as in "nuclear deterrence" escalation. The last word is matched as you type it, so modern already finds modernization, and the search box suggests completions drawn from the words in the quotes themselves. The top completion appears greyed out after what you typed: press Tab or → to take it. Use ↑ and ↓ to step through the suggestions, and each one's results load as you reach it. Press Enter or pick a suggestion to search right away; a new search always replaces the one before it. Choosing Highlight, Summaries or Blocks under Search in narrows the match to those parts of a card.
On a phone
The list of quotes is the whole screen. Tap a quote to focus it: a Back button returns to the list exactly where you left it, and three tabs sit under it — Quote (the card), AI summary (the card's saved find-flaws-and-extensions analysis when one exists — opening a quote never starts a new one; press Generate for that) and Full page (the source article, extracted through qwksearch). Every quote you open has all three. On a wide screen the list, card and AI panel stay side by side as before.
Only a newly generated analysis counts toward your plan's daily card AI analyses: loading a saved one is free, and an analysis that fails is not counted.
Read the source article beside a card
When a card's citation (or its body) links the web page it was cut from, two buttons appear under the citation:
- Open page opens the original article in a new tab.
- Full article pulls the article's full text and a formatted citation through qwksearch's article extractor and shows it beside the card — side by side on a wide screen, below the card on a narrow one. The highlighter is on by default: select any passage to mark it, tap a mark to remove just that passage, the eraser clears all your marks, the highlighter button turns marking off and on, and the copy button puts the citation plus your highlighted passages (or the whole article, if you haven't marked anything) on the clipboard, ready to cut a new card.
Some pages can't be extracted — paywalled articles, PDFs and sites that block crawlers. The panel says so and offers Retry or Open page instead. The article closes when you select a card from a different source.
1. The Debate Singularity is Happening
AI now researches topics in depth, highlights key quotes, frames both sides of arguments, coaches preparation, exposes flaws in evidence cards, and crafts strategic closing speeches that compare competing claims. Integrating AI into debate unlocks the next stage of emergent complexity in socio-political governance.
2. Collective Thought Engine
Language Models can distill collective thought into a vector space where every point carries a weighted value, reflecting its contribution to decision-making. This collective AI consciousness can synthesize complex arguments, judge their validity and relevance, and support a democratic, AI-mediated economy where public votes reward influence and insight.
AI agents learn which arguments and sources persuade different audiences. Crowdsourced automation enables AI to organize argument outlines—much like GitHub's reusable code model—helping it evaluate complex decisions with greater depth and consistency.
3. Transparent Reasoning
AI reveals the exact sentences and citations behind its reasoning, allowing users to verify alignment with collective interests through sentence-by-sentence interpretability.
Debate Garden aggregates perspectives across the ideological spectrum, highlighting the most persuasive arguments on every side of an issue. As public trust in media and institutional expertise has declined, achieving genuine neutrality requires presenting all viewpoints — not just the mainstream consensus. Rather than offering reductive caricatures of opposing positions, this platform equips users with the strongest, most rigorously reasoned arguments available.
4. Outcome Simulation Trees
Users can engage in multiple AI-powered practice rounds that map trees of possible outcomes and responses, surfacing the most persuasive strategies across diverse contexts.
5. Outlines of Current News Issues
Each debate card fits into a broader narrative—a topic tree linking evidence, assumptions, and values within a collective framework. This transformation shifts debate from isolated rounds to a shared model of research crowdsourced to the public by a few thousand editors, like Wikipedia or Github.
6. Solving Post-Self Alignment
Recognizing the interconnectedness of research articles mirrors how we should relate to global citizens as part of an emergent collective consciousness. Losing sight of that bigger picture fuels bias, tribalism, and division at the root of modern conflicts between isolated social news bubbles. Only by mapping the whole internet as a debate outline enables seeing how each idea fits into a whole greater than the sum of its parts. This sets a socio-political model to build synergistic consciousness ethical awareness across both discourse and institutional governance. This can reduce LLM hallucination and steer alignment with common social values as AI gains capacity to replace human leaders of organizations.
7. Topic Research Unified Tree Hierarchy (TRUTH)
Key metadata extracted for LLM analysis allows models to detect logic flaws, flag overstatements, strengthen warrants, and place each claim within the Topic Research Unified Tree Hierarchy (TRUTH). This consensus-driven system seeks grounded truth, reduces hallucination, and aligns AI reasoning with shared human values—laying groundwork for responsible AI governance.
