Wireless peripherals
Wireless keyboards and mice were in ordinary use by 2009. He was right about the technology, but less accurate about consumer preference: wired devices remained in use.
Read the evidenceKurzweil put deadlines on wireless networks, wearable computers and self-driving cars. We checked 84 of his forecasts against what happened.
Networks and medical tools fare better in this collection than robotics and transport. Of 16 computing forecasts, 10 came true, 5 by the deadline.
68 predictions with deadlines passed · Our judgment
Arrived by the deadline.
e.g. Wireless peripheralsCame true, behind schedule.
e.g. Brain-scale supercomputingPartly fulfilled by review.
e.g. Robots in every homeStill unfulfilled at review.
e.g. Life expectancy over 100Our judgments on a selected collection, not a lifetime accuracy score. 16 other records are future or undated. Scope & sources
In our default reading, 3 of 4 network forecasts and 4 of 5 medical forecasts were fully realized by the review date. Robotics and transport: 1 of 5. These are small groups drawn from an incomplete collection.
68 predictions with deadlines passed
Open a field to see its predictions. Hover a square for the claim; select it for the evidence.
Fully realized = on time + late. Partial gets no fractional credit. Both percentages use the assessed predictions in that field. Sample sizes and deadlines differ between fields.
Claims about computing capacity, size and cost.
Transport and availability of information, not language understanding.
Software abilities. Tag commercial adoption separately.
Distinguish headsets, retinal displays, contact lenses and direct neural interfaces.
Separate computational research milestones from human clinical outcomes.
Keep population, age group, access and life-expectancy definition explicit.
Brain hardware, brain understanding and cognition augmentation.
Distinguish prototype, commercial service and widespread use.
Medical nanobots: primary category Biology & medicine; nanotech as a tag.
Preserve energy versus electricity, capacity versus generation.
Track geography and population; separate causal claims from observed events.
Do not confuse availability with dominant use.
Distinguish device existence from prevalence or displacement of personnel.
The two largest groups target 2009 and 2019: 61 of these 68 forecasts. Each column shows the outcomes for one target year, assessed through October 15, 2026.
Select a year to see its predictions. 68 predictions with deadlines passed.
Target year · one column per group
Older targets have had longer to catch up, and each group contains different claims. Only due forecasts are shown; year columns are equally spaced.
84 predictions · 68 with deadlines passed · Our judgment
Capability asks whether a technology exists; adoption asks how widely people use it.
| Prediction type | Due | On time | Late | Partial | Missed |
|---|---|---|---|---|---|
| Adoption | 34 | 7 | 6 | 12 | 9 |
| Capability9 future | 25 | 14 | 4 | 4 | 3 |
| Performance / cost1 future | 3 | 0 | 1 | 0 | 2 |
| Research milestone | 2 | 2 | 0 | 0 | 0 |
| Social / institutional4 future · 2 undated | 2 | 0 | 0 | 0 | 2 |
| Trend | 2 | 2 | 0 | 0 | 0 |
14 future and 2 undated predictions are outside the outcome counts. Select a count to explore its predictions in the ledger.
We judge each claim against its deadline and intended scale. A working product can satisfy a capability forecast; a prediction of widespread use also needs evidence of adoption.
Consumer preference and regulation can limit use even when a technology works. We judge those cases individually and show alternative readings where the result is ambiguous.
Wireless keyboards and mice were in ordinary use by 2009. He was right about the technology, but less accurate about consumer preference: wired devices remained in use.
Read the evidenceTianhe-2 exceeded the 20-petaflop benchmark in 2013, four years after the deadline. This measures computing performance, not a simulation of the human brain.
Read the evidenceRobot vacuums handle a specific chore. The forecast called for robots in nearly every household, doing a broad range of chores.
Read the evidenceSame 67 predictions in every reading. One brain-modeling forecast is left out because its alternative deadline has not passed.
| Reading | Predictions | On time | Fully realized |
|---|---|---|---|
| Reference (stated reasonable reading) | 67 | 37.3%25 / 67 | 53.7%36 / 67 |
| Strict / alternative reading | 67 | 25.4%17 / 67 | 37.3%25 / 67 |
| Broader / directional reading | 67 | 55.2%37 / 67 | 92.5%62 / 67 |
The main comparison has 68 due predictions. This comparison uses the 67 eligible in all three readings. Broader success may mean meeting different conditions; open a record to see what changed.
Each record includes the forecast, our judgment, source links and earlier assessments. Open one to check the reasoning or compare alternative readings.
Explore the predictionsIt compares 84 distinct forecast versions. Repeated claims count once; materially different deadlines or thresholds stay separate. 68 are due, 14 are future, and 2 lack a usable deadline.
The selection is incomplete and uneven across fields. Most book claims were checked through reprints or retrospectives; original editions still need checking. The research and judgments are AI-assisted and have not had independent human review.
Outcomes are assessed through October 15, 2026. The sources and downloads include all 88 original intake records and earlier assessments.