Product
Three outputs, one goal: you always know what's going on in tech. Built from 9 communities, processed daily.
Intelligence Reports
CxO, AI, Developer, Content: each report reads the same underlying data through a different lens. TL;DR, key events with numbers, deep analysis, and cited sources. Available daily, weekly, monthly, or quarterly. Full archive so you can go back to any period and understand what you missed.
Read a report →Keyword Intelligence
Two years of tech discourse, structured and queryable. Track how any company, tool, person, or idea has moved over time: sentiment, volume, and when the narrative shifted. Whether something is genuine momentum or small chatter. Source links all the way back to the original discussion.
See trending keywords →API Access
Five endpoints covering keywords, time-series graphs, sentiment, connected topics, and AI summaries. Query any range from daily to 2 years. Every result linked back to its original source in the tech conversation.
API docs →Not search results. Not press releases. The communities where tools get adopted, companies get criticized, procurement decisions get questioned, and ideas spread before they go mainstream. Hacker News, Reddit, GitHub, X, Substack, Discord, YouTube, ArXiv, where the actual tech conversation happens.
What makes the reports trustworthy is the data underneath them. Instead of summarizing headlines, we process raw tech discourse through purpose-built models. The same pipeline has been running and improving since 2023.
01
Daily crawls across Hacker News, Reddit, GitHub, X, Substack, Discord, YouTube, ArXiv, and RSS feeds, the places where tech gets built, debated, and decided.
02
A custom fine-tuned keyword extraction model identifies tech terms: companies, tools, people, and concepts, from up to 20,000 texts per day.
03
A sentiment model scores each text. A categorization model then assigns each keyword one of 12 category labels. Both run in batches on a dedicated GPU.
04
Keywords, sentiment, engagement, and source links are aggregated into clean, queryable data. Long-term storage in BigQuery. API access via MongoDB. Updates daily at 6 AM UTC.
General-purpose models hallucinate on niche tech content. Smaller models fine-tuned on domain-specific data outperform them at a fraction of the cost.
Keyword extractor
Custom fine-tuned model that identifies tech terms: tools, companies, people, and concepts, from raw crawled text. Trained specifically on tech discourse, not general language. Open-sourced on Hugging Face.
Sentiment analyzer
Scores each text as positive, negative, or neutral. Applied per-text in batches of 20–40 at a time. Processes 10,000+ texts in under 30 minutes on a single T4 GPU.
Categorization model
A RoBERTa-based model fine-tuned to classify each keyword across 12 labels: companies, tools, platforms, AI models, and more. Trained on keyword-plus-context, not the keyword alone. 99% accuracy on held-out data.
Five endpoints. Six time periods. Every keyword is linked to its original sources, so you can drill from trend to full context in one call. Public endpoints are free with rate limits. Private endpoints (longer time ranges, richer data) require an API key.
Read the docs →/v2/keywordsTop and trending keywords for a given time period and category.
/v2/graphTime-series data for a keyword: count, sentiment, and engagement over time.
/v2/connected-keywordsKeywords that frequently co-occur with a given keyword.
/v2/sourcesOriginal source URLs and excerpts for a keyword within a time range.
/v2/ai-summaryAI-generated summary for a keyword and period: includes a full narrative, short summary, and one-line description.
The reports and keyword explorer are live. Start there. For private API access, longer time ranges, and full historical data, apply for access and we'll be in touch.