Writings
Posts on technology, organizations, strategy, systems, and the people operating within them.
Exploring how systems behave in practice. How decisions get made, where technology helps or fails, how incentives shape outcomes, and what happens when technical and organizational realities collide.
AI Reliability: Why Context Capacity and Verification Capacity Must Grow Together
Reliable AI depends on relevant context and the capacity to verify its outputs. Retrieval, source traceability, independent checks, monitoring, and correction need to develop together.
AI Governance: How Organizations Manage AI Risk, Accountability, and Responsible Use
AI governance connects policies, ownership, risk assessment, testing, and monitoring so organisations can make and revisit accountable decisions about AI throughout its lifecycle.
AI Integration: How to Connect AI With Existing Business Systems
AI integration connects models to business data, applications, and workflows. Reliable operation depends on access controls, data quality, human review, recovery, and ongoing maintenance.
AI Tools: What They Are, How They Work, and How to Choose the Right One
AI tools span generation, prediction, analysis, and automation. Choosing one requires representative evaluation, clear data boundaries, and a workflow that supports reliable use.
Bugs as Breadcrumbs for Better UX
Bug reports can reveal where an interface leaves people uncertain, loses their work, or makes recovery difficult. Following those clues helps teams improve the experience around a failure as well as correct its technical cause.
Debugging and Thinking: Every Error Tells a Story
Effective debugging turns an unexpected result into a sequence of observations, hypotheses, and discriminating experiments. A useful explanation accounts for the failure, guides a focused repair, and makes the behaviour easier to recognise next time.
AI-DLC: What Changes When AI Becomes Part of the Development Lifecycle
AI-DLC brings AI into requirements, design, construction, and operations. Its practical value depends on how well teams preserve context, make decisions, and verify the work that moves between those stages.
Social Loafing: Understanding Uneven Contribution in Teams
Social loafing describes reduced individual effort during collective work. Addressing an uneven workload requires understanding what people were expected to contribute, what prevented progress, and how the team recognises and supports shared work.
The Power of Framing in Conflict Resolution
Workplace conflict often hardens around the first explanation a team accepts. The way a disagreement is framed shapes whose perspective gets heard, where accountability lands, and what the organisation tries to fix.
The Circuit Breaker Pattern: Containing Failure in Distributed Systems
Circuit breakers limit repeated calls to a failing dependency. Their value depends on sensible timeouts, meaningful failure signals, honest fallback behaviour, and a recovery path that has been tested.
Contract Testing vs Integration Testing: What Each Can Tell You
Contract tests check the expectations at service boundaries, while integration tests exercise connected components. A useful testing strategy gives each a clear purpose and checks the versions that will actually run together.
Legacy System Modernisation: Choosing What to Change and How to Get There
Legacy system modernisation starts with the constraints the business needs to remove. Choosing an approach requires understanding existing behaviour, data ownership, operational responsibilities, and the route to retiring what is replaced.
Mock Servers: Building Against an API That Does Not Exist Yet
Mock servers let teams develop against an agreed API before the provider is ready. Their usefulness depends on realistic scenarios, repeatable tests, and checks that keep the simulation aligned with the actual service.
Technical Debt: Why Moving Fast Today Can Slow Software Tomorrow
Technical debt makes later changes more expensive when a system's structure no longer supports the work required of it. Managing it means identifying recurring friction, making trade-offs explicit, and choosing improvements with a concrete purpose.
Role Ambiguity: What It Is and Why It Matters
Role ambiguity can make reasonable decisions look like poor performance. When ownership, priorities, and authority remain implicit, people end up working towards different versions of the same job.
The Latency of Human Thought in Autonomous AI Systems
An AI system can produce an answer faster than a person can evaluate it. Useful interfaces need to support reading, verification, and decisions as carefully as they support generation.
Understanding Risk Aversion and Why We Fear Failure
Risk aversion is often treated as an immutable personality trait. In reality, it is frequently a rational response to the system, incentives, and transparency people operate within.
Rest as an Engineering Primitive: Designing Work That Allows Recovery
Sustained overtime changes the conditions under which software decisions are made. Rest, realistic workloads, and reliable handovers belong in engineering planning alongside testing, review, and recovery.
Cognitive Reappraisal: How Teams Manage Stress Better
Cognitive reappraisal involves reconsidering what a difficult situation means. At work, it can help people examine assumptions and choose a response, provided it remains grounded in evidence and does not disguise problems that need changing.
Local-First Architecture and the Moral Dimension of Data Sovereignty
Local-first software lets people continue working with data on their own devices while supporting synchronisation and collaboration. Its value depends on more than offline access, including usable exports, recovery, and clear limits on what the architecture guarantees.
Cloud Security and the Human Decisions Behind It
Cloud security depends on the decisions people can realistically make under pressure. Authentication, access controls, and reporting processes need to account for interruptions, uncertainty, and the ordinary demands of getting work done.
The Dashboard Fallacy: Why More Metrics Lead to Less Understanding
When an operations team monitors 120 live charts across multiple wall displays, they are rarely diagnosing faster—they are desperately filtering noise from signal.
The Quiet Workspace: Reducing Visual and Auditory Churn
A quiet workspace depends on more than muted notifications. It needs clear priorities, usable controls, and team expectations that let people concentrate without losing access to important information.