AI Coding Bots as Systemic Risk
Draft
Why speed without model discipline can leak failure modes into legacy systems, internal tools, and shared infrastructure.
- Local optimizers are dangerous when the blast radius is enterprise-wide.
- Agent supervision is still a human capital problem.
- Tooling adoption often precedes governance by several quarters.
Leverage Is Usually the Story
Draft
How constraints, financing structures, and optionality explain market behavior better than most commentary.
- High beta demand is often balance-sheet demand in disguise.
- Structured liquidity changes founder behavior long before cap tables show it.
- Narratives are downstream of funding conditions.
Cycles Humble Every New Cohort
Outline
Each generation rediscovers drawdowns, acquisition ego, loose diligence, and the cost of mistaking momentum for competence.
- The wrappers change; the behaviors barely do.
- Declining trust forces deeper diligence everywhere.
- Most market wisdom looks obvious only in arrears.
Polyphasic Supervision for Agents
Idea
A thought experiment on AI oversight, asynchronous review loops, and what happens when software has no single sleep cycle.
- Continuous operation changes the cadence of human review.
- Escalation paths matter more than raw model quality.
- Monitoring becomes a product, not a task.
