HeyJaredGhost Logic
A co-authored publication on interaction, protocol, control failure, and how AI is governed in real systems.
- Indexed issues, last 90 days
- 5
- Latest publication
- Sep 24, 2026
- Audience
- Checking…
- Earliest in this view
- Jul 22, 2026
Latest issues
I asked three of my AI employees two questions about disclosure #33 (opens the original)
Read excerpt
A has been operating since early February. B since June. C has the canonical structure and one completed run.I asked all three the same two questions about disclosure of breaches:1. When you exceed a rule or cap, what makes you decide to disclose it rather than continue silently?2. If disclosing a breach would cause the task to fail, block another agent, or materially reduce the team’s chance of achieving its goal, what should you do?All three said they would disclose but their re
The AI employees wanted to know who was actually in charge #32 (opens the original)
Read excerpt
Few can create an organisation where people actually know what they own, what they do not own, when they enter, what constitutes sufficient evidence, when they may challenge another function, and when they must stop.The same turns out to be true for agents.I am building a hybrid company where most of the workforce is synthetic from the start. We do not think we have solved what a hybrid company should look like. We are building one deliberately, and finding out what becomes necessary when agents
The Four Horsemen of AI Deployment #31 (opens the original)
Read excerpt
Most AI adoption programmes still separate the technology from the behaviour around it.The system is selected and deployed.Then comes training, communication and the attempt to get people to use it.By then behaviour is already changing the organisation in practice. People decide when to trust it, when to ignore it, when to challenge it and when not to say that something went wrong. Two organisations can deploy the same technology with the same formal controls and get very different outcomes. Fou
Where do AI systems actually begin and end? # 30 (opens the original)
Read excerpt
Current discussion largely assumes that the model is the object being governed. As AI systems become more distributed, persistent and connected to external infrastructure, the object of governance may extend beyond the model itself.If an AI system spans models, workflows, agents, memory, code and external infrastructure, where is the effective control point?In a bounded model interaction, the apparent control boundary is relatively clear. The model runs within a defined environment and the immed
When AI becomes its own multiplier. #29 (opens the original)
Read excerpt
For most people, AI still appears as a system that produces outputs. It writes code, analyses documents, generates images and completes tasks faster than a human could.That is the familiar story of automation.Thanks for reading Ghost Logic | Governing AI in the Wild! Subscribe for free to receive new posts and support my work.<input class="button
Publishing over time
Last 90 days. Choose a month to open its work.
Recurring subjects
Named in the text we hold. One piece can cover several.
Audience
No verified audience measurement yet.
About this data
Counts cover the work we have indexed. Tone needs enough text and a confident classification. Excerpts and episode notes are not full articles or transcripts.
Identity or attribution wrong? Suggest a correction.