KNEW v2 — Production-capable. Live. Scaling.
Since the EIC submission, KNEW has progressed from validated prototype to a production-capable global news intelligence platform. We redesigned ingestion and the AI pipeline, expanded coverage and languages, and hardened the system for enterprise and EU-wide deployment.
KNEW is no longer a feasibility experiment — it is a working platform delivering value to early users. We moved the program from building a prototype to scaling reliability and reducing unit cost to enable broad adoption.
Product maturity
Full end-to-end engine: hybrid ingestion (RSS + crawlers + conditional APIs), 8-layer AI processing (bias, sentiment, fact-check, summarization, categorization), geocoding & ownership mapping, and a live UI in production.
Architecture & reliability
Rebuilt ingestion to a fault-tolerant, async worker queue with Redis-backed caching & TTLs, automated retries, and observability (metrics, logs, SLI/SLOs) for production stability.
Cost & scale
New hybrid strategy reduces per-article inference cost by an order of magnitude: RSS + local crawlers for coverage, conditional API calls for high-value sources, and caching to avoid repeated inference.
- Hybrid ingestion: RSS + lightweight crawlers + conditional APIs.
- Async AI pipeline: bias detection, neutral summarization, credibility scoring, sentiment, fact-checking (worker queues + Redis caching).
- Language-aware preprocessing and geospatial mapping for multi-country coverage.
- Observability: metrics, error alerts, retry policies; production logging & secure secrets management.
Early users
~100 active testers with daily/weekly usage; qualitative feedback indicates clarity & relevance improvements over generic aggregators.
Willingness to pay
High intent signals for a professional subscription (~€9/month) among power users (journalists, analysts).
Evidence
Proposal PDFs (Part B sections) are attached in the archive for audit: download or request the full docs in the interview.
Scale ingestion
Engineer hires for crawler & ops, regional collectors in 25+ countries, and scalable collectors (3–6 months).
AI & cost-efficiency
Self-hosted inference pilots, model distillation, caching & batching to reduce inference cost and improve latency.
Enterprise & compliance
GDPR, auditability, SLAs, and enterprise API work to onboard newsrooms, institutions, and research partners.
0–3 months
Crawler expansion pilots, finalize self-hosted inference PoC, observability & SRE hardening.
3–9 months
Scale to 25+ countries, pilot enterprise integrations, tighten compliance processes.
9–12 months
Launch paid tiers, onboard first enterprise customers, begin EU-wide deployments.
Current
~100 early users · Live ingestion & AI pipeline · Daily test traffic.
Targets
Latency < 2s for priority feeds · Cost per 1k articles ↓ 70% · 10k DAU in 12 months.
Impact
Reduced misinformation exposure, faster situational awareness for journalists & institutions, improved media transparency across regions.