Your AI Bill Doubles on 1 December. Has Anyone Told Finance?
Reading Time: 5 minutesGemini 4 Argon costs $2/$10 per million tokens until 30 Nov, then $4/$20. See every tier, one workload re-priced, and five steps before 1 December.
Your Boss’s AI Can Still Watch You Work. It Can’t Guess How You Feel
Reading Time: 5 minutesCalifornia’s AB 1883 bars AI that infers workers’ emotions or collects neural data. See what it bans, what stays legal, and the safety clause lawyers dispute.
One Request, Every Data Broker: What California’s DROP Actually Deletes
Reading Time: 5 minutesCalifornia DROP lets residents send one deletion request to 600+ data brokers. Here is what brokers must do, what it skips, and what the first fine was for.
The Token Diaries #4 — Your Prompts, Their Data
Reading Time: 5 minutesAI skill pays, but every prompt is also a log entry. What employees should know about prompt records, shadow AI and who reads your AI diary.
Ecosystem Restoration
Reading Time: 3 minutesAI ecosystem restoration gone wrong? A thought experiment on AI side effects, using the 1935 cane toad release as a warning.
The Token Diaries #3 — Tokenmaxxing Inc.
Reading Time: 6 minutesLeaderboards, mandates and a 50-year-old law: what happens when AI usage becomes the scoreboard, and what companies should count instead.
India’s Consent Managers Open on 13 November. The Office That Registers Them Is Not Staffed
Reading Time: 5 minutesDPDP consent manager registration starts about 13 November 2026, but the Board that takes applications was reported not yet functional.
Resource Competition
Reading Time: 3 minutesAI resource competition, explained as a thought experiment: DeepMind’s apple game, AI pricing agents and the RealPage antitrust case.
The Token Diaries #2 — The Developer’s Ledger
Reading Time: 5 minutesInput, output, cached: three meters run every AI call. How caching, model routing and context tiers cut a developer’s token bill without hurting quality.
Production Diaries #1 — Can three free models beat one?
Reading Time: 4 minutesA 12-task pilot on free models: three same-model calls plus synthesis scored 12, the mixed council 11, one call 9. Inconclusive, and here is why.