Category
07 essays
AI essays
LLM product decisions, agents, evals, and shipping AI features teams can own.
AI
8 min read
AI MVP in 2026: What "Minimum" Means When the Model Is the Risk
A practical framework for scoping an AI-powered first release when the model, the API, or the data pipeline is the real unknown, not the interface.
AI
7 min read
How to Choose an LLM for a Product in 2026
A decision matrix for picking a model class by latency, cost, context length, tool use, multimodality, data residency, and lock-in, plus why the choice needs a quarterly revisit.
AI
7 min read
RAG vs Fine-Tuning vs Agents: Pick the First Release Shape
A comparison of retrieval-augmented generation, fine-tuning, and agentic architectures, with a default decision path for choosing the right shape for a v1.
AI
7 min read
Agentic AI in Products: Workflow, Not Chatbot Theater
Why agentic AI succeeds as a bounded workflow with permissions, audit logs, and human checkpoints, not as an open-ended chatbot with tool access.
AI
8 min read
Long-Context LLMs (Kimi and Peers): When 100k+ Tokens Actually Help
A practical framework for deciding when a 100k-1M+ token context window earns its cost, and when retrieval still wins on price, freshness, and cacheability.
AI
7 min read
AI Product Ownership: What to Hand Off After the First Demo
The operational package a team needs after an AI feature clears its first demo: runbooks, eval suites, prompt and version ownership, cost dashboards, and on-call coverage.