Mop the slop from your LinkedIn feed with a new open-source Chrome extension

Tom Frazier launched Slop Mop, a free open-source Chrome extension that uses TypeSafe's Jev probabilistic decision model to flag or hide low-value LinkedIn posts, whether AI- or human-written. The tool analyzes up to 250 posts per user per day at an early estimated cost of roughly $0.005 per user daily, while LinkedIn AI-content concerns remain elevated: Pangram classified 41% of sampled posts exceeding 250 words as fully AI-generated. The project provides an early practical showcase for Jev's API-based, confidence-scored decisioning model, though its current commercial impact is limited.

Analysis

The investable signal is not content filtering itself but evidence that low-cost, probabilistic routing models may cannibalize a portion of higher-cost LLM inference in enterprise workflows. The highest-value use cases are binary or scored decisions—triage, moderation, fraud-review prioritization, document classification and workflow escalation—where latency and cost matter more than generative quality. This is directionally negative for inference-revenue assumptions embedded in premium-model providers, but it is too early to underwrite a material near-term impact without independently verified pricing, throughput, retention, and enterprise deployment data.

For platform operators, widespread user-side filtering is a warning that feed-quality degradation can become an engagement and advertiser-yield issue. LINKEDIN is not separately investable within Microsoft (MSFT), so the near-term read-through is immaterial to consolidated earnings; however, a durable shift toward quality-ranking tools raises the strategic value of proprietary interaction data and credible provenance signals. Meta (META), Alphabet (GOOGL) and Reddit (RDDT) are more exposed over 6-18 months if AI-generated engagement bait reduces trust, time spent, or creator economics.

Contrarian view: the market may overstate the implication for frontier LLM demand. Cheap classifiers can reduce calls to generative models, but they also create more automated workflows that ultimately require generation, retrieval, audit trails, and human escalation. The key gating issue is false-positive tolerance: if filtering suppresses legitimate posts or creates reputational risk, consumer adoption will remain niche and enterprise moderation buyers will favor incumbent trust-and-safety stacks rather than open-source extensions.

Key Decisions for Investors

• No standalone trade on this launch; monitor TypeSafe/Jev adoption for disclosed API pricing, enterprise customers, daily request growth and gross-margin evidence over the next 1-3 months before assigning public-market read-through.
• Maintain a 6-18 month quality-risk watch on RDDT versus META: RDDT's monetization is more dependent on perceived authenticity and community quality, while META has superior ranking data and moderation scale. Reassess if reported time spent, creator retention, or ad-load guidance weakens alongside rising AI-content complaints.
• For AI infrastructure exposure, favor diversified hyperscalers MSFT and GOOGL over a concentrated premium-inference thesis until lower-cost decision-model substitution rates are measurable. The thesis is falsified if enterprise workloads demonstrate that classifier adoption increases, rather than displaces, total frontier-model token consumption.
• Set an alert for platform policy changes requiring AI-content labels or provenance metadata. A broad regulatory or app-store mandate would benefit identity, content-safety and governance vendors more than consumer AI-content detection tools, but no targeted position is warranted without named public beneficiaries and revenue disclosures.