bradshaw.cloud: Audio Edition
by John Bradshaw
A/B Testing Models: Your Model Choice Is a Hypothesis, Not a Decision
About this podcast
Essays and strategy insights on cloud economics, edge architecture and AI infrastructure, read aloud in a synthetic clone of the author’s voice. Each episode is the full text of a piece from bradshaw.cloud, narrated by a synthetic voice cloned from John Bradshaw with his consent; the written version, with its diagrams and tables, is linked from every episode. Personal views, not those of my employer.
Episodes
The newest 20 of 55 episodes
- A/B Testing Models: Your Model Choice Is a Hypothesis, Not a Decision · 9 min · Season 2026 Episode 51Controlled experiments are routine for interface changes and rare for the model behind a product. With model prices and capability moving every few weeks, an unchallenged…
- Where Inference Runs Is a Cost Decision, and the Crossover Runs Both Ways · 12 min · Season 2026 Episode 50Hourly accelerator prices are the wrong unit for deciding where self-hosted inference runs. Measured as cost per useful answer, central placement wins for large, batchable work…
- Managed or Self-Run · 6 min · Season 2026 Episode 49The premium on a managed database, queue or search service buys an on-call rota. Self-running open source wins only when the fleet is large enough to spread that fixed cost.
- Model Routing: The Cheapest Model Is the One You Didn't Call · 11 min · Season 2026 Episode 48The largest inference saving is rarely a cheaper model. It is sending each request to the smallest model that can answer it well, and governing that routing decision as a…
- Your Internal Developer Platform Will Not Survive AI Unchanged · 11 min · Season 2026 Episode 45AI changes the job of an internal developer platform. Model access, data context, evaluation, agent runtime, cost attribution and policy now need a supported path, not six teams…
- A PWA Is Not an Edge Strategy · 5 min · Season 2024 Episode 1How Progressive Web Apps leverage distributed and edge cloud services for faster load times, offline resilience, and improved user experiences.
- Typed Decision Models · 13 min · Season 2026 Episode 42The class of models that decide instead of chatting — why a typed probability in milliseconds is worth more, and costs less, than a frontier paragraph.
- Most AI Requests Aren't AI Problems · 8 min · Season 2026 Episode 47Many AI requests are process-ownership and decision-rights problems that arrive as tool requests. A five-question test to run before funding one, and where to route a request…
- Data Residency Is Not Sovereign AI · 18 min · Season 2026 Episode 32A local model region answers one question about location. Sovereign AI also requires control of prompts, embeddings, model provenance, operations, jurisdiction and a tested exit…
- AI Platform Operating Model · 5 min · Season 2026 Episode 31AI does not replace platform engineering. It expands the platform’s job from delivery pipelines to model access, data context, evaluation and agent governance.
- Agent Authority · 5 min · Season 2026 Episode 30An AI agent needs more than an API token. The framework for assigning a sponsor, mandate, permissions, autonomy and evidence before it can act.
- Agentic API Security · 16 min · Season 2026 Episode 34The caller is no longer a person. The identity, task-scoping, reversibility gates, and observability an API estate needs once autonomous agents hold the credentials.
- Team Topology Decisions · 11 min · Season 2026 Episode 28Team design is an architectural act, not an HR one. The four team types and three interaction modes from Team Topologies — plus the decision layer: when each shape fits, and what…
- AI Delegation Loop · 5 min · Season 2026 Episode 40Stop building a specialist agent for every job. Capture process, toolbox and proof in a playbook a general agent can run — and put every correction back into the folder.
- An AI Agent Is Not a Service Account · 20 min · Season 2026 Episode 46An AI agent can act with or without a human in the loop. That needs more than a credential: a named sponsor, bounded mandate, scoped permissions, explicit autonomy and an…
- Technical Debt Prioritisation · 11 min · Season 2026 Episode 29Not all debt is worth repaying, and the messiest code is rarely the most expensive. A prioritisation framework built on interest rate, change frequency, blast radius, and…
- The Model Context Protocol: A Strategic Reference for Technical Leaders · 14 min · Season 2026 Episode 20MCP standardises how LLM apps call tools, moving integration from M×N towards M+N. It does not make a connection safe — identity, scopes and review stay yours.
- Sovereign AI · 5 min · Season 2026 Episode 33Data residency is one control in a larger AI sovereignty model. Score data, models, operations, jurisdiction and exit paths before you choose a deployment.
- Cloud Repatriation · 5 min · Season 2026 Episode 19A decision framework for 2026 — the real survey data, a workload scoring matrix, a 3–5 year TCO model, and the risks. Repatriation as portfolio management, not ideology.
- FinOps for AI · 8 min · Season 2026 Episode 26AI spend scales with autonomy, not traffic. The circuit breakers, tier routing, caching, and chargeback discipline that keep LLM and agent spend answerable to somebody.








