Dr. Kalun Ho

Videos

Every video, with its handout, deck and sources.

Newest first. Each entry has the one-sentence takeaway, the free PDF handout, the editable PowerPoint deck and the sources, numbered as in the video.

Video 1 · November 2026 · Explainer

AI Got 280× Cheaper. So Why Is Your Bill Going Up?

Your AI bill is set by tokens per task, not price per token: budget per task and route each task to the smallest model that does it.

Watch on YouTube Handout and deck (PDF and PowerPoint, Google Drive) The experiment
Sources
  1. Stanford HAI (2025). The 2025 AI Index Report. Takeaway 7: GPT-3.5-level inference cost fell over 280-fold, Nov 2022 to Oct 2024. hai.stanford.edu/ai-index/2025-ai-index-report. Accessed 8 November 2026.
  2. Cottier, B.; Snodin, B.; Owen, D.; Adamczewski, T., Epoch AI (2025). LLM inference prices have fallen rapidly but unequally across tasks. Cheapest model above a benchmark threshold, USD per 1M tokens blended 3:1. epoch.ai/data-insights/llm-inference-price-trends. Accessed 8 November 2026.
  3. Pichai, S., Google (2026). I/O 2026: Welcome to the agentic Gemini era. Tokens processed per month across Google's surfaces; company-reported. blog.google/innovation-and-ai/sundar-pichai-io-2026. Accessed 8 November 2026.
  4. Tully, T.; Redfern, J.; Das, D.; Xiao, D., Menlo Ventures (2025). 2025: The State of Generative AI in the Enterprise. US enterprises only; survey of 495 decision-makers plus a market model. menlovc.com/perspective/2025-the-state-of-generative-ai-in-the-enterprise. Accessed 8 November 2026.
  5. Eurostat (2025). Use of artificial intelligence in enterprises (isoc_eb_ai). Enterprises with 10+ employees; Germany from the dataset (updated 2026-06-15). ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises. Accessed 8 November 2026.
  6. OpenAI (2026). API pricing. Standard tier, USD per 1M tokens. developers.openai.com/api/docs/pricing. Accessed 8 November 2026.
  7. Anthropic (2026). Pricing. Base input and output, USD per 1M tokens. platform.claude.com/docs/en/about-claude/pricing. Accessed 8 November 2026.
  8. Google (2026). Gemini API pricing. Standard tier; Gemini 4 Argon price from 2026-12-01 as listed on the page. aistudio.google.com/docs/pricing. Accessed 8 November 2026.
  9. Hadfield, J.; Zhang, B.; Lien, K.; Scholz, F.; Fox, J.; Ford, D., Anthropic (2025). How we built our multi-agent research system. anthropic.com/engineering/multi-agent-research-system. Accessed 8 November 2026.
  10. Gundlach, H.; Lynch, J.; Mertens, M.; Thompson, N., MIT FutureTech (2026). The Price of Progress: Price Performance and the Future of AI (arXiv:2511.23455). arxiv.org/abs/2511.23455. Accessed 8 November 2026.
  11. Dr. Kalun Ho (2026). Cheaper Tokens, Costlier Tasks: Measuring Tokens per Task Across Reasoning Settings, Model Size and Agentic Tool Use. Own measurement; setup, full results and code on the page. kalunho.com/experiments/tokens-per-task. Accessed 8 November 2026.
  12. Qwen Team, Alibaba (2026). Qwen3.8-27B model card. reasoning_effort low, medium, xhigh; xhigh is the default. huggingface.co/Qwen/Qwen3.8-27B. Accessed 8 November 2026.
  13. Jevons, W. S. (1866). The Coal Question (2nd ed.; first published 1865). Chapter VII, Of the Economy of Fuel. econlib.org/library/YPDBooks/Jevons/jvnCQ.html?chapter_num=9. Accessed 8 November 2026.
  14. European Central Bank (2026). Euro foreign exchange reference rates (USD). 2026-11-06, 1 EUR = 1.1165 USD. ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/index.en.html. Accessed 8 November 2026.
  15. Ong, I.; et al., UC Berkeley Sky Computing Lab (2024). RouteLLM: Learning to Route LLMs with Preference Data (arXiv:2406.18665). sky.cs.berkeley.edu/project/routellm. Accessed 8 November 2026.
  16. Chen, L.; Zaharia, M.; Zou, J. (2023). FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performance. arxiv.org/abs/2305.05176. Accessed 8 November 2026.

The handouts and decks are stored on Google Drive. The link opens Google Drive in a new page; nothing loads from Google while you only read this page. How linked services handle your data