What AI has actually done, with the receipts

As of September 2026, AI has done these measurable things: AI-supported screening found 29% more breast cancers in a randomised trial of over 100,000 women, an AI-discovered drug improved lung function in a Phase 2a trial where placebo declined, AI flood forecasts reach 100 countries where 700 million people live, and 49% of U.S. adults now use AI chatbots. Every claim here names who measured it, when, and the date we checked it, with the receipts downloadable below.

The record by area

Seven results, with sources

  1. AI-supported mammography found 29% more cancers with no increase in false positives

    29% more cancers detected; 44% less screen-reading workload

    Source: The Lancet, via EurekAlert, 2026-01-29. Checked 2026-09-06.

  2. The first AI-discovered drug to report a randomised Phase 2a result improved lung function where placebo declined

    FVC +98.4 mL (60 mg daily) vs -20.3 mL (placebo) at 12 weeks

    Source: Nature Medicine, via EurekAlert, 2025-06-03. Checked 2026-09-06.

  3. The 2024 Nobel Prizes in Chemistry and Physics went to AI work

    Chemistry: protein structure prediction and design; Physics: neural networks

    Source: The Nobel Foundation, 2024-10-09. Checked 2026-09-06.

  4. AI river-flood forecasts cover 100 countries where 700 million people live, seven days ahead

    100 countries; 700 million people; 7-day lead time

    Source: Google, 2024-11-11. Checked 2026-09-06.

  5. California's AI camera network detected 636 wildfires in 2024 before anyone called 911

    1,668 fires picked up by cameras; 636 (38%) before any 911 call; 1,211 cameras

    Source: Gov1 (quoting CAL FIRE), 2025-09-24. Checked 2026-09-06.

  6. More than 1 million blind or low-vision people use Be My Eyes, with AI answering alongside 10 million volunteers

    1,000,000+ users; 10,000,000+ volunteers; 150+ countries; 180+ languages

    Source: Be My Eyes, 2026-03-12. Checked 2026-09-06.

  7. Half of U.S. adults now use AI chatbots, and about a quarter use them daily

    49% use AI chatbots (33% in 2024); ~24% daily; ChatGPT 44%

    Source: Pew Research Center, 2026-06-17. Checked 2026-09-06.

Why this site exists

Most of what people read about AI is a forecast, a fear or a pitch. This site records only what has already happened and been measured by someone with no stake in this site: journals, regulators, meteorological centres, survey organisations. Where the record is good, it says so with the number. Where a result is early, it says that too: one Phase 2a trial is a start, not a cure. The methodology page lists what is deliberately left out.

Frequently Asked Questions

Measured, published results as of September 2026 include: a randomised trial of over 100,000 women in which AI-supported mammography found 29% more cancers with no rise in false positives; the first AI-discovered drug to improve lung function in a randomised Phase 2a trial; an AI database of over 214 million protein structures used by more than 4.5 million people, recognised with the 2024 Nobel Prize in Chemistry; AI flood forecasts covering 100 countries where 700 million people live; and AI cameras that spotted 636 California wildfires in 2024 before anyone called 911. Each has a primary source on this site.

It is pro-evidence. The ledger records only things that have happened and been measured or certified, with the source that did the measuring. It does not record forecasts, and it does not argue that harms do not exist; the methodology page says what is left out and why.

Yes. Every entry is available as CSV and JSON from the data links on every page, under a CC BY 4.0 license, so you can check, quote and reuse it with attribution.

Through Amazon affiliate links to two books by Ray Kurzweil and to Audible memberships. The commission is paid by Amazon, never changes what you pay, and has no bearing on which results appear in the ledger.

Every entry carries the date the fact refers to and the date it was checked against its source, most recently 2026-09-06. Entries are rechecked and the check date updated; an entry that fails a recheck is removed.