An algorithm is desperately needed though. Sources have varying publishing rates and quality. Aggregating across them and culling is crucial. For the most part I've thrown up my hands and delegated this job to hacker news and Reddit subscriptions, but it's still not well personalized.