AI-First Work Requires Organizational Redesign
The durable advantage will come from redesigning workflows, roles, and incentives around AI—not simply adding tools to existing processes.
- future of work
- organizational design
- AI strategy
Writing archive
Blog
Writing on AI, product strategy, software systems, and the future of work.
Where Specialized Decision Models May Beat General LLMs
A small customer-conversation evaluation suggests focused decision models may offer a useful cost and latency advantage for narrow classification tasks.
- AI evaluation
- conversational AI
- product experiments
Direct AI with Positive Instructions
Clear targets usually guide a language model better than a context window filled with the behaviours it should avoid.
- AI
- prompting
- productivity
Curiosity Is Getting Cheaper
New creative AI tools reduce the cost of turning a passing question into an experiment, making exploration faster and more playful.
- creative AI
- experimentation
- product thinking
A Simple Long-Term Memory for AI Coding Agents
Two Markdown layers can preserve durable project knowledge across coding-agent sessions without turning memory into noise.
- AI agents
- software engineering
- developer tools
Data Moats in the Age of World Models
Long-running, hard-to-reproduce datasets can become a durable advantage when new AI capabilities need a faithful picture of the real world.
- AI strategy
- data
- world models
Manual vs. Automated: The MVP Decision That Actually Matters
Keep uncertain workflows manual until customer value and repeated behaviour show the team what is worth automating.
- MVP
- product management
- startups
Vintage LLMs for Historical Storytelling
A period-specific model could help storytellers recover the assumptions, details, and language that modern models smooth away.
- AI
- storytelling
- experiments
Measure the Human Touches in Your Team
A small leadership metric for making appreciation deliberate instead of accidental.
- leadership
- teams
- management
From Pure RAG to Breadcrumbs and Self-Healing Retrieval
Why retrieval works better as a navigation system than as a larger pile of disconnected text chunks.
- AI
- RAG
- system architecture
Let LLMs Think; Let Software Do
Reliable AI systems separate probabilistic reasoning from deterministic state, rules, data routing, formatting, and execution.
- software architecture
- LLMOps
- AI agents
From Product Thinking to Product Building
AI prototyping tools are moving product managers from static artifacts toward functional prototypes that can be tested before engineering scales them.
- product management
- prototyping
- AI tools
AI's HTML Moment: Can MCP Become a True Open Standard?
MCP may solve AI's interoperability problem, but technical openness, ecosystem effects, and neutral governance will determine whether it becomes lasting infrastructure.
- AI infrastructure
- MCP
- open standards
LLMs Are Cheap Now. Where Is the Dishwasher?
A batching analogy for finding efficiency in products that make large numbers of language-model calls.
- AI
- sustainability
- product management
2025 was the Land Grab. 2026 is the Reality Check.
AI infrastructure is moving from aggressive capital allocation to the harder work of operating complex systems profitably.
- AI infrastructure
- operations
- product strategy