101 · Where to start

From anxiety to action, in five movements.

1. 01

   The premise, once.

   The world will not lack AI. It will lack the capacity to use it well. A surgeon does not operate in ignorance of complications. Naming them is what makes the operation possible — and trustworthy.

2. 02

   Proportion is a form of truthfulness.

   Every concern is graded. Severity is the scale and reversibility of harm. Likelihood is how often it is already observed. Their product is a composite score from 1 to 25. A nuisance is not presented with the same weight as an irreversible harm.

3. 03

   The map has a grain.

   Three parts, 23 domains, 116 families, 1000 named concerns. Click from the whole to a single entry in a few steps. Color follows the part; size follows activity and score.

4. 04

   Questions before verdicts.

   Each entry carries starter questions written for curiosity and learning, not accusation — and a principle (trustworthiness, listening, hope, justice, service) to sit beside the technical fact.

5. 05

   One thing this week.

   Open the map. Choose the branch that touches your work or your family. Read the definition. Hold one question with someone else. Return when the news changes — the register is built to be lived in, not finished.

Six concerns the field is already living with

• Cross-context prompt injection

  Attacker exploits shared context between users or sessions.

• Many-shot jailbreaks

  Exploiting long contexts to shift model behavior.

• Autonomous vulnerability discovery

  Agents scanning codebases for exploitable flaws.

• Autonomous exploit development

  LLMs writing working exploits from CVE descriptions.

• AI-augmented malware

  Polymorphic, target-adaptive malware.

• Password and credential-stuffing acceleration

  AI generating plausible password guesses.