Long-form journalism, reviewed before it ships
Each article passes a ten-agent AI review and a weighted human review before it appears. The bar is accuracy, not engagement.
How review works
Every long-form article enters a multi-step pipeline before it can publish. A claim-extractor pulls factual statements out of the draft. A fact-verifier checks them against external sources. A source-assessor weighs each citation. A bias-detector flags one-sided framing. A rhetoric-analyzer looks for manipulative language. Three more agents look at objectivity, AI-generated content, and omissions. A readability scorer rates the prose. Articles that clear the pipeline cleanly publish automatically. Articles flagged for any reason go to human reviewers — domain-qualified contributors who score the work without seeing the author's name. Their scores are aggregated with weights based on track record and expertise.
Breaking news, without the cost of being wrong
Verified and independent journalists can publish breaking stories immediately, marked Developing Story, while verification continues in the background. The deadline is 24, 48, or 72 hours depending on the topic's risk profile. If verification fails, the article is marked Verification Failed and stays online with the failure displayed — corrections happen in public, not in silence. Authors can revise and resubmit.
Corrections, appeals, and the public record
Readers can propose corrections. Authors and editors can issue them. Each correction is tied to the specific claim it addresses, the version history is public, and the diff is visible to anyone. Reputation moves with the record: contributors earn standing through accurate work and lose it through repeated correction. Authors who think a moderation decision is wrong can appeal, and appeals are processed in a queue with the same blind-review framing as initial AI review.
Why this is different
Most platforms optimise for engagement and rate-limit corrections after the fact. Ekta inverts that. The cost of publishing wrong is built into the workflow, and the reputation system rewards contributors who have been right before. There is no algorithmic reach lift for outrage. There is no boost for engagement velocity. The signal is verifiability.