The source is part of the answer.
Collected 2026-09-23T14:26:27+00:00. These pages contain public source records, not illustrative data.
Software support
We combine the endoflife.date API v1 responses for eight selected database and language products. We retain release dates and reported maintenance status, count records by status, and link each row to its product page. This community-maintained source can differ from vendor policies; follow its source links before making support decisions.
API documentation · MIT license and copyright notice
Known exploited vulnerabilities
We collect CISA’s complete KEV JSON catalog, verify its declared record count, preserve CVE identifiers, and count records by vendor label. We do not infer that a particular system is vulnerable or rank vendors by safety. We link each record to CISA for current guidance. DatumHorizon is independent and is not endorsed by CISA.
Original JSON feed · CC0 license and source terms
AI model releases and performance
The AI models page is a dated, selected research snapshot assembled from public provider documentation and official model cards. We preserve model/version, release precision, API availability, open-weight status, license wording, context, price, latency, throughput, and benchmark settings as separate fields. Provider-reported benchmarks are grouped by exact benchmark/version label but are not treated as independent reproductions or a universal ranking. Artificial Analysis was reviewed but excluded from page data because its terms restrict republishing extracted scores, serving measurements, and price data; missing evidence remains explicit.
AI model snapshot · source ledger and artifact checksum · Chart.js MIT license
Dungeon Crawler Carl signals
Thirteen monthly Wikimedia article-view totals reconcile to 396 daily records. We chart September 2025 through August 2026, compare monthly totals with daily averages, and retain August 2025 for a same-month annual comparison. These are Wikipedia views, not Reddit posts or unique readers. Reported sales, crowdfunding and community snapshots retain their separate units and dates.
Recent NYC construction activity
We query the official DOB NOW filings and approved-permit datasets for the latest three source calendar dates, reconcile the declared count before normalization, and join BBL values to PLUTO in bounded batches for property context. We also collect six complete months of issuance counts grouped by month and work type, reconciled against a separate source count. The page preserves separate source-update and collection timestamps, leaves missing joins unknown, and excludes applicant or owner contact fields. DOB electrical, elevator, and LAA job families are published separately.
DOB NOW filings · DOB NOW approved permits · PLUTO property context · NYC Open Data
Frontier model thinking levels
We compare reasoning levels, benchmark evaluations, token pricing, and latency across OpenAI, Google, and Anthropic frontier models. The data normalizes intelligence indices across competition mathematics, expert scientific reasoning, and software engineering benchmarks, correlating capability gains with task costs and completion speeds.
Thinking levels comparison · source ledger and artifact checksum
Residential electricity bills
Five annual EIA Table 5A files cover 2020–2024, 50 states, DC and the published national total. We reconcile customer counts and weighted averages, retain source-file checksums, and distinguish price per kWh from consumption and bills. Context comes from a separate EIA explanatory publication and an Ipsos survey. These are nominal historical averages, not tariffs or current household bills.
Electricity bills and comparisons
Vehicle safety recalls and defect investigations
We analyze 35 top-selling US consumer vehicle models across model years 2020 through 2025 using the NHTSA Recalls API, NHTSA Safety Complaints database, and official NCAP/IIHS crash safety evaluations. We quantify recall severity, distinguish urgent Park Outside fire advisories and Do Not Drive orders, track resolution channels (over-the-air software updates versus dealer hardware repairs), and correlate component defect complaint patterns with official crash ratings.
Vehicle safety recalls · source ledger and artifact checksum
Vehicle reliability and longevity
We evaluate 35 benchmark consumer vehicle models across passenger cars, pickup trucks, SUVs, and electrified powertrains using long-term odometer survival studies from iSeeCars (analyzing 395M+ vehicle transactions), three-year initial quality and problem frequency from J.D. Power (PP100), and national fleet age data from the Bureau of Transportation Statistics and S&P Global Mobility. We correlate high-mileage survival rates with component simplicity, powertrain architecture (naturally aspirated V6/V8 and planetary e-CVTs versus downsized turbos and dual-clutch transmissions), and lifetime maintenance expenditures.
Vehicle reliability study · source ledger and artifact checksum
Airline flight delays, cancellations, and carrier reliability
We analyze 15 reporting U.S. air carrier networks and major regional partners using an eight-source authoritative ledger including the Bureau of Transportation Statistics (BTS TranStats), the Department of Transportation Air Travel Consumer Report (ATCR), the DOT Airline Customer Service Commitments Dashboard, GAO operational disruption investigations (GAO-23-105524), FAA ATCSCC airspace flow management data, Cirium aviation telemetry benchmarks, MIT ICAT delay propagation models, and 14 CFR Part 260 refund standards. We track on-time gate arrival rates within 15 minutes of schedule (A15 standard), isolate late-arriving aircraft propagation from controllable air carrier maintenance delays, measure passenger complaint rates per 100,000 enplanements, and explain mandatory passenger cash refund rights.
Airline reliability study · source ledger and artifact checksum
Food safety, grocery, and pet food recalls
We analyze 35 landmark recall campaigns across FDA CFSAN, USDA FSIS, and FDA CVM regulatory jurisdictions. We quantify statutory hazard severity under 21 CFR Part 7 (Class I life-threatening hazards versus Class II moderate risks), track biological pathogen outbreaks (Listeria monocytogenes, Salmonella enterica, E. coli STEC), isolate physical foreign materials and undeclared allergens, cross-reference confirmed CDC PulseNet hospitalization data, and explain consumer refrigerator sanitization and retailer refund rights.
Food safety recalls · source ledger and artifact checksum
Collection and publication
Refreshes run when we deploy or manually start the repository workflow; there is currently no automatic daily refresh. A complete validated snapshot is stored in DynamoDB and read back to generate these pages. A failed collection or validation keeps the last published pages available. The displayed time is collection time, not a claim that upstream data changed then.
Records are deduplicated by product/release, CVE identifier, or source dataset identifier. Missing lifecycle dates remain explicitly unknown. Counts describe source coverage, not the entire software market. Recent-window charts describe that window. NYC historical charts use separate, explicitly dated aggregates from the current published permit dataset.