AI

IndustrySpace

Overview Community Topics Debate claims News Sources Projects People Papers Ontology Claims map Governance Activity

Vote duration

24h

Pass threshold

51%

Universal threshold

100%

Fast pass threshold

Quorum

News:MiniMax releases M3.1 Flash AI model at $0.15 per million tokens with major performance gains

Voting period open

News:AI Could Push High-Risk ER Patients Down the Queue, Study Warns

Voting period open

News:Meta creates new class of non-voting shares to fund AI investments

Voting period open

Add missing years to AI policy, defense and security claims

Voting period open

Correct OpenAI funding-round month and add years to funding claims

Voting period open

Publish the Debate argument sourcing AI workflow with its six stages

Voting period open

Edit debate relations for "Nikola Danaylov vs. Ash Kumra on Large-scale retraining programs can successfully mitigate AI-driven job displacement". Added: Topics — Art, Philosophy, Literature, Creative writing, Creative thinking, Automation, Apple iPad, Education system, Enterprise AI adoption, AI in learning, Divergent thinking, Workforce training, Technological Unemployment, Forecasting, Humanities, Artificial intelligence, AI Models, AI-driven layoffs; Countries — Norway, Finland...

Voting period open

TEst notifications

Voting period open

APEC finance ministers reach consensus on AI and digital financing in Hong Kong

100%

15h 9m remaining

Cloudflare releases Clairic extraction model on Workers AI

100%

15h 10m remaining

News:Cybercriminals bypass AI guardrails in hacking campaigns as models refuse to help defenders

100%

15h 10m remaining

News:Government confirms indefinite recognition of CE marked medical devices

100%

15h 10m remaining

News:GUARD-LGG AI model predicts radiotherapy response in lower-grade glioma patients

100%

15h 11m remaining

News:AI chatbots recognize unnecessary medical tests but struggle where doctors disagree, study finds

100%

15h 11m remaining

News:Allen Institute for AI releases FairStress benchmark for testing LLM fairness

100%

15h 11m remaining

News:Study finds most AI incidents involving U.S. companies not disclosed in SEC filings

100%

15h 12m remaining

News:Curl project shifts low-severity vulnerabilities to public disclosure due to report overload

100%

15h 12m remaining

Loading initial...

Review proposal

Publish the Debate argument sourcing AI workflow with its six stages

Ali Raza

Fast pathVoting period open

Verify each find in code, not by the model

Current

Proposed edits

Verify each find in code, not by the model

Workflow stage

Description

Checks that the article belongs to this space, that the quote matches a claim Geo extracted or appears word for word on the page, and that any web outlet carries the space's accepted tag. A find that fails any check is blocked and cannot be approved.

Order

Input

Candidate sources and quotes from the previous stage.

Output

Finds marked verified or blocked, each with the check that decided it.

Tools

Debate Desk

Models

Claude Sonnet 5.5

Operator roles

Automated

Review

Sync debates and rebuild spoken order

Current

Proposed edits

Sync debates and rebuild spoken order

Workflow stage

Description

Pulls Debate entities and their transcripts from Geo's debate service and rebuilds them into the order the words were actually said, labelling each round. Geo's stored transcript blocks are not in spoken order, so presenting them as the sequence would misrepresent the debate.

Order

Input

Target Geo space and its recorded Debate entities.

Output

Speaker turns in spoken order with round labels, and the arguments extracted from each.

Tools

Claude Code

Debate Desk

Models

Claude Sonnet 5.5

Operator roles

Automated

Debate argument sourcing with human approval

Current

Proposed edits

Debate argument sourcing with human approval

AI workflow

Description

Finds a published story or report that supports or contradicts each spoken argument in a Geo debate, verifies every quote in code rather than trusting the model, and publishes only what a curator approves. Runs on local Claude Code, so no stage costs anything.

Input

A Geo space and its recorded debates, with the arguments spoken in them.

Goals

Back every spoken debate argument with a source outside the debate

Publish accurate AI curated entities to Geo

Workflow stages

Sync debates and rebuild spoken order

Index the stories a space already accepts

Find evidence for one spoken argument

Verify each find in code, not by the model

Approve, shorten or reject every find

Publish approved sources to the space

Tools

Claude Code

Debate Desk

Geo AI

Models

Claude Sonnet 5.5

Claude Opus 5.5

Status

Active

Maintainers

Ali Raza

Providers

Anthropic

Topics

AI-assisted knowledge curation

Debate Desk

Current

Proposed edits

Debate Desk

Tool

Description

Local dashboard where AI finds evidence for arguments made in Geo debates and a curator approves every item before it is published. Runs the model through the operator's own Claude Code rather than an API key.

Approve, shorten or reject every find

Current

Proposed edits

Approve, shorten or reject every find

Workflow stage

Description

Every verified find needs a curator's ruling before it can go anywhere. Editing may only shorten a quote, never reword it, so an approved quote is always a substring of what the source actually says.

Order

Input

Verified finds awaiting a decision.

Output

Approved finds, with rejected and blocked ones recorded and never published.

Tools

Debate Desk

Models

Claude Sonnet 5.5

Operator roles

Human-in-loop

Review

Back every spoken debate argument with a source outside the debate

Current

Proposed edits

Back every spoken debate argument with a source outside the debate

Goal

Description

Arguments made in recorded Geo debates cite only the debate itself. Link each one to a published story or report that supports or contradicts it, drawn from sources the space accepts.

Index the stories a space already accepts

Current

Proposed edits

Index the stories a space already accepts

Workflow stage

Description

Copies every Article in the space, plus the claims Geo itself extracted from each one, into a local full-text index. Searching what Geo already holds before going to the web is what keeps the evidence inside sources the space accepts.

Order

Input

A Geo space, its Articles and the claims extracted from them.

Output

Searchable index of accepted stories and their real body sentences.

Tools

Claude Code

Debate Desk

Geo AI

Models

Claude Sonnet 5.5

Operator roles

Automated

Find evidence for one spoken argument

Current

Proposed edits

Find evidence for one spoken argument

Workflow stage

Description

Searches the indexed stories first and only then the open web, requiring at least six distinct searches before it may report that Geo holds nothing. Support or opposition is decided from the body of the source, never the headline.

Order

Input

One spoken argument from a debate, and the space it belongs to.

Output

Candidate sources with the exact sentence that supports or contradicts the argument.

Tools

Claude Code

Debate Desk

Models

Claude Sonnet 5.5

Claude Opus 5.5

Operator roles

Automated

Publish approved sources to the space

Current

Proposed edits

Publish approved sources to the space

Workflow stage

Description

Turns approved finds into Sources relations, creating an Article and its Publisher only where Geo has none. Runs the ported safety gates — additions only, every new entity typed, no values on a relation id — and the curator types publish to send one proposal per space.

Order

Input

Approved finds for one Geo space.

Output

One governance proposal per space linking each spoken argument to its source.

Tools

Debate Desk

Geo AI

Models

Claude Sonnet 5.5

Operator roles

Manual

Review