How LLMs Predict the Next Word
45sUses a simple analogy to explain the core mechanism of LLMs, making a complex topic instantly understandable.
▶ Play Clip"The title accurately reflects the content: a concise explanation of large language models."
Snapshot 2026-11-17
Every observation TubeSum recorded, plotted as measured.
The vertical axis starts at 6.8M, not zero, so later observations stay readable.
Based on 80 observations · last observed Nov 17, 2026 · solid lines connect consecutive measurements (≤2 days apart); dashed lines span days TubeSum did not measure. · vertical axis focused on the observed range (starts at 6.8M)
Exact daily observations; gaps mean no observation was taken that day.
| Date | Views | Avg views / hour | Channel median |
|---|---|---|---|
| 2026-11-17 | 8,600,764 | 493.2 | — |
| 2026-11-16 | 8,595,416 | 493.1 | — |
| 2026-11-15 | 8,572,728 | 492.7 | — |
| 2026-11-14 | 8,560,887 | 492.6 | — |
| 2026-11-13 | 8,539,064 | 492.3 | — |
| 2026-11-12 | 8,527,100 | 492.0 | — |
| 2026-11-10 | 8,489,651 | 491.2 | — |
| 2026-11-09 | 8,470,279 | 490.7 | — |
| 2026-11-08 | 8,449,648 | 490.3 | — |
| 2026-11-07 | 8,437,141 | 490.2 | — |
| 2026-11-06 | 8,421,726 | 490.0 | — |
| 2026-11-05 | 8,405,687 | 489.8 | — |
| 2026-11-04 | 8,387,995 | 489.4 | — |
| 2026-11-03 | 8,362,056 | 489.0 | — |
| 2026-11-02 | 8,355,932 | 488.9 | — |
| 2026-11-01 | 8,338,253 | 488.6 | — |
| 2026-10-31 | 8,327,705 | 488.7 | — |
| 2026-10-30 | 8,315,062 | 488.6 | — |
| 2026-10-29 | 8,299,117 | 488.3 | — |
| 2026-10-28 | 8,281,523 | 488.0 | — |
| 2026-10-27 | 8,265,064 | 487.7 | — |
| 2026-10-26 | 8,245,709 | 487.5 | — |
| 2026-10-25 | 8,229,744 | 487.3 | — |
| 2026-10-24 | 8,223,731 | 487.4 | — |
| 2026-10-23 | 8,205,182 | 487.3 | — |
| 2026-10-22 | 8,198,139 | 487.2 | — |
| 2026-10-21 | 8,176,521 | 487.0 | — |
| 2026-10-20 | 8,162,958 | 486.8 | — |
| 2026-10-19 | 8,151,857 | 486.7 | — |
| 2026-10-18 | 8,137,216 | 486.5 | — |
| 2026-10-17 | 8,127,779 | 486.5 | — |
| 2026-10-16 | 8,114,043 | 486.4 | — |
| 2026-10-15 | 8,093,392 | 486.2 | — |
| 2026-10-14 | 8,084,837 | 486.1 | — |
| 2026-10-13 | 8,061,757 | 485.7 | — |
| 2026-10-12 | 8,056,694 | 485.6 | — |
| 2026-10-11 | 8,028,915 | 485.1 | — |
| 2026-10-10 | 8,015,975 | 485.0 | — |
| 2026-10-09 | 8,002,376 | 484.9 | — |
| 2026-10-08 | 7,993,756 | 484.8 | — |
| 2026-10-07 | 7,975,692 | 484.4 | — |
| 2026-10-06 | 7,952,572 | 483.9 | — |
| 2026-10-05 | 7,945,942 | 483.9 | — |
| 2026-10-04 | 7,915,972 | 483.2 | — |
| 2026-10-03 | 7,906,314 | 483.0 | — |
| 2026-10-02 | 7,891,489 | 482.8 | — |
| 2026-10-01 | 7,864,994 | 482.2 | — |
| 2026-09-30 | 7,842,550 | 481.5 | — |
| 2026-09-29 | 7,818,685 | 480.8 | — |
| 2026-09-28 | 7,796,043 | 480.1 | — |
| 2026-09-27 | 7,777,697 | 479.7 | — |
| 2026-09-26 | 7,761,099 | 479.3 | — |
| 2026-09-25 | 7,744,538 | 479.0 | — |
| 2026-09-24 | 7,723,475 | 478.4 | — |
| 2026-09-23 | 7,703,198 | 477.9 | — |
| 2026-09-22 | 7,679,456 | 477.1 | — |
| 2026-09-21 | 7,666,263 | 476.8 | — |
| 2026-09-20 | 7,639,717 | 476.1 | — |
| 2026-09-19 | 7,627,015 | 476.0 | — |
| 2026-09-18 | 7,610,859 | 475.5 | — |
| 2026-09-17 | 7,585,491 | 474.6 | — |
| 2026-09-16 | 7,545,931 | 473.1 | — |
| 2026-09-15 | 7,524,363 | 472.4 | — |
| 2026-09-14 | 7,505,723 | 472.0 | — |
| 2026-09-13 | 7,488,818 | 471.6 | — |
| 2026-09-12 | 7,479,439 | 471.5 | — |
| 2026-09-11 | 7,464,434 | 471.3 | — |
| 2026-09-07 | 7,389,992 | 469.5 | — |
| 2026-09-04 | 7,348,425 | 469.1 | — |
| 2026-09-01 | 7,307,470 | 468.5 | — |
| 2026-08-27 | 7,230,318 | 467.3 | — |
| 2026-08-17 | 7,132,472 | 468.2 | 1,262,848 |
| 2026-08-10 | 7,061,168 | 468.8 | 1,262,848 |
| 2026-08-05 | 7,021,388 | 469.8 | 1,262,848 |
| 2026-07-31 | 6,990,822 | 471.6 | 1,262,848 |
| 2026-07-27 | 6,966,149 | 472.9 | 1,262,848 |
| 2026-07-24 | 6,944,938 | 473.8 | 1,262,848 |
| 2026-07-23 | 6,935,368 | 474.0 | 1,262,848 |
| 2026-07-22 | 6,928,264 | 474.4 | 1,262,848 |
| 2026-07-20 | 6,917,301 | 475.0 | 1,262,848 |
Audience reaction — separate from the performance metrics above.
Introduction to LLMs as Next-Word Predictors
Building a Chatbot and Adding Randomness
Training Data and Parameters
Training Process and Computational Scale
Reinforcement Learning from Human Feedback
Hardware and Transformer Architecture
Attention and Feed-Forward Networks
Emergent Behavior and Conclusion