Build an evidence-led newsroom.
PSYTECH News is a live publication that makes psychology central to reporting on AI, neuroscience and technology. Its product system distinguishes discovering sources from assessing evidence and deciding what deserves publication.
Aggregation is not editorial judgement.
Research claims arrive through press releases, preprints, independent journalism and institutions with different evidence standards. Scaling feeds without source judgement can reproduce unverified claims, duplicates or shallow psychology commentary. The challenge is to grow global discovery without lowering publication standards.
Source breadth with explicit boundaries.
- Multilingual discovery. Configure feeds from research organisations, journals, regulatory bodies and professional reporting with original language and attribution retained.
- Provenance-aware relevance. Translate before a second English relevance gate; preserve the original source and disclose translation.
- Differentiate source authority. Treat preprints as provisional, corporate assertions as first-party claims and university press communications as summaries rather than independent replication.
- Fail-closed editorial checks. Implement deduplication, relevance, source validation and a requirement to make the psychological mechanism defensible.
- Readable publication design. Lead a restrained editorial visual system, source notes and mobile navigation rather than a generic auto-generated feed.
Documented scale—not invented reach.
These are system configuration counts from merged code, not audience numbers, completed daily ingestion volume, or independent editorial approval. The public news site is live, but automated newsroom jobs have recently failed.
Translate an editorial point of view into a system.
I defined the psychology-first proposition, product strategy, editorial standards, quality gates and the visual reading experience, coordinating AI-assisted implementation. This integrates commercial product thinking, editorial judgement and behavioural science. Technical source configuration is not presented as proven business impact.
Publishing reliability must be demonstrated.
Recent automated newsroom and GitHub CI runs are failing. A further editorial redesign is still in a draft pull request, and historical records show gaps in affirmative human approval evidence. We have not yet demonstrated reliable end-to-end editorial sign-off, audience growth, revenue or sustained autonomous publication.
Evidence boundary
All screenshots are from the public site. Private editorial controls, credentials, unpublished branch features and confidential source code are excluded. Continue to the Clinical Reasoning Desk case study.