First snapshot (news lane, 2026-09-24)

README.md

@@ -0,0 +1,71 @@
+---
+license: cc-by-4.0
+pretty_name: 'Keyfern: What People Search About Superintelligence (Sept 2026)'
+language:
+- en
+tags:
+- keyword-research
+- autocomplete
+- search-intent
+- seo
+- artificial-intelligence
+- superintelligence
+- ai-policy
+- tabular
+- cybermax
+task_categories:
+- text-classification
+- tabular-classification
+size_categories:
+- 1K<n<10K
+configs:
+- config_name: default
+  data_files:
+  - split: train
+    path: data/*.parquet
+---
+
+# Keyfern: what people search about superintelligence (Sept 2026)
+
+<!-- cybermax-more:start -->
+> **Try the paid tool:** [Keyfern on Apify](https://apify.com/cybermax/keyword-ideas): get long-tail keyword ideas and questions for your own seeds from Google, YouTube, Bing, Amazon and DuckDuckGo autocomplete. First try costs cents; pay only for results.
+<!-- cybermax-more:end -->
+
+"Superintelligence" moved from books and sci-fi into the news in the week of 22 Sep 2026: an AI remark in a UN General Assembly speech (22 Sep) and a proposed US bill to ban artificial superintelligence (reported by AP on 23 Sep; see the [Policyloom AI policy tracker](https://huggingface.co/datasets/shaw276/policyloom-us-ai-policy-tracker) for the verified events). This dataset records **what people were typing into search boxes about it that week**: **1,106 unique autocomplete suggestions** for four seeds (`superintelligence`, `super intelligence`, `artificial superintelligence`, `ai ban`), collected on 24 Sep 2026 from Google, YouTube and Bing autocomplete (588 lookups: seed, seed + a–z, 12 question words, 10 prepositions).
+
+Each row has the suggestion, the seed that produced it, an intent label (question, comparison, commercial, transactional, local, informational; a heuristic from the wording), a topic cluster, which engines suggested it and how often.
+
+**Quick facts [Verified, from the rows]:** 115 suggestions are questions (e.g. "is superintelligence possible", "what is superintelligence") and 27 are comparisons (e.g. "superintelligence vs agi"). "superintelligence ban" and "ai ban news" show up on all three engines. A large cluster is the "artificial superintelligence alliance" crypto token, not the policy topic, so filter by `cluster` before drawing conclusions.
+
+Autocomplete shows **what people type, not search volume**. No volumes are invented here. For direction over time, check Google Trends (see below).
+
+## Use it
+
+```python
+from datasets import load_dataset
+df = load_dataset("shaw276/superintelligence-search-questions", split="train").to_pandas()
+print(df[df.intent == "question"].sort_values("times_suggested", ascending=False)[["keyword", "engines"]].head(20))
+```
+
+## Columns
+
+| column | meaning |
+|---|---|
+| `snapshot_date` | collection date (UTC) |
+| `seed` | the seed phrase that produced the suggestion |
+| `keyword` | the autocomplete suggestion (lower-cased) |
+| `intent` | heuristic intent from the wording |
+| `cluster` | first new word the suggestion adds to the seed |
+| `engines`, `engine_count` | which of google / youtube / bing suggested it |
+| `times_suggested`, `best_position` | how many lookups returned it; its best position in a suggestion list (1 = top) |
+| `found_via` | the first lookup that found it (`seed`, `alphabet:x`, `question:how`, `preposition:vs` …) |
+
+## Get fresh data
+
+- **Keyword ideas and questions for your own seeds, 5 engines, with intent, clusters and new-keyword alerts:** [Keyfern on the Apify Store](https://apify.com/cybermax/keyword-ideas).
+- **Is interest rising or falling?** [Swellmeter: Google Trends scraper and API](https://apify.com/cybermax/google-trends) gives interest over time with a plain-English rising/falling verdict.
+- Free: [rising searches Space](https://huggingface.co/spaces/shaw276/rising-searches) · [what people search Space](https://huggingface.co/spaces/shaw276/what-people-search) · [US AI Policy Tracker](https://huggingface.co/spaces/shaw276/us-ai-policy-tracker).
+
+## Notes
+
+Raw search suggestions: they can include people's names, brands and products because searchers type them. They are shown as collected, with no editing or endorsement. US English market. Neutral and factual; not affiliated with Google, Microsoft or YouTube. By CyberMax.

data/snapshot-2026-09-24.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22eb1eb32cff70c5109d1df2e99cb109e70ece71e96b6401c379628cc524eafd
+size 25399