The AI Caramba journal
Explore together
how AI + Data Security evolves.
Questions, experiments and personal tool picks.
Two formats: Explorations and Top 3.
Exploring AI Data Security · Part 3
An AI-generated summary is a new document with a new destination. Part 3 shows how labels, usage rights and DLP must follow the information beyond the source file.
Read the article →Latest articles
13 articlesRead from the beginning →Editorial archive dates · August–September 2026
- Data Security
Exploring AI Data Security · Part 2
A shared file can carry facts and hostile instructions through the same context. Part 2 follows that path from document ingestion to a tool action and places a control at every boundary.
- Explorations
Exploring Mid-Session Changes · Part 4
What can an agent do when work gets stuck? We compare more reasoning, expert advice, parallel agents and a model switch.
- Explorations
Exploring Mid-Session Changes · Part 3
How do retries and human work change the cost of a model switch? We compare token savings, correction time and the cost of failed attempts.
- Explorations
Exploring Mid-Session Changes · Part 2
How conversation state, prompt caching and compaction shape the cost of a running AI session—and how to weigh a smaller context against rebuilding a reusable cache.
- Explorations
Exploring Mid-Session Changes · Part 1
When does switching to a cheaper model save money? We compare the next call, the remaining calls and the cost of handing over the work.
- Top 3
My top 3 tools for multi-repository projects
Beads for tasks, Matt Pocock’s skills for implementation and OpenWiki for context. The tools I want to keep when I switch coding environments—plus a minimal AGENTS.md to take away.
- AI
GPT-6 Astra: from clever answers to sustained work
Astra improves long-context retrieval and task execution. Its new Codex memory design and better handling of boundaries point toward a more useful kind of autonomy.
- AI
Your AI knows the code. Who knows the plan?
Self-awareness, context handling and coordination: how capable agents keep track of the task and build on earlier work.
Follow the journal.
Follow every published article in your own RSS reader.