M · thread
AI / ML lifecycle
If and how AI features go in the product: build-on-APIs vs. fine-tune vs. skip, model and provider choice, evals, cost control, and the lifecycle of shipped AI behavior.
19 decision topics in this thread — every topic gives a recommended default, the reasoning, and what flips it. Titles and dates below are free; the verdicts and full topics are the paid masterclass (a few showcase verdicts are shown in full).
| ID | Topic | Verdict | Verified |
|---|---|---|---|
| AI-001 | Is ML the right tool? (problem framing) | �e9f6e; verdict in the full masterclass | 2026-06-26 |
| AI-002 | Build-vs-buy: hosted API vs. open-weights vs. fine-tuned | For your first shipped product and the first 12–18 months: use a hosted API. | 2026-06-26 |
| AI-003 | Model selection & benchmarking | �e9f6e; verdict in the full masterclass | 2026-06-26 |
| AI-004 | Inference hosting & serving | �e9f6e; verdict in the full masterclass | 2026-06-26 |
| AI-005 | AI-native feature & product decisions[A] | �e9f6e; verdict in the full masterclass | 2026-06-26 |
| AI-006 | On-device vs. cloud inference2026 | �e9f6e; verdict in the full masterclass | 2026-06-26 |
| AI-007 | Prompt design & versioning | �e9f6e; verdict in the full masterclass | 2026-06-26 |
| AI-008 | RAG: vector store, embeddings, chunking | �e9f6e; verdict in the full masterclass | 2026-06-26 |
| AI-009 | Agentic features: tool-calling, MCP & orchestration2026 | �e9f6e; verdict in the full masterclass | 2026-06-26 |
| AI-010 | Fine-tuning / distillation / adapters | �e9f6e; verdict in the full masterclass | 2026-06-26 |
| AI-011 | Training-data sourcing & governance | �e9f6e; verdict in the full masterclass | 2026-06-20 |
| AI-012 | Evaluation harness | �e9f6e; verdict in the full masterclass | 2026-06-26 |
| AI-013 | Safety guardrails | �e9f6e; verdict in the full masterclass | 2026-06-26 |
| AI-014 | Hallucination management & grounding | �e9f6e; verdict in the full masterclass | 2026-06-26 |
| AI-015 | AI content provenance & marking2026 | �e9f6e; verdict in the full masterclass | 2026-06-26 |
| AI-016 | Cost, token budgeting, caching & fallback2026 | �e9f6e; verdict in the full masterclass | 2026-06-26 |
| AI-017 | AI observability & drift monitoring | �e9f6e; verdict in the full masterclass | 2026-06-26 |
| AI-018 | Human-in-the-loop & escalation | �e9f6e; verdict in the full masterclass | 2026-06-26 |
| AI-019 | Model governance & responsible-AI review | �e9f6e; verdict in the full masterclass | 2026-06-26 |
Each topic behind these titles is a full worked decision: what it is, the recommended default, why, what flips it, how to execute it with your coding agent — cited and dated.
Read a full topic free →Launching soon — read a free chapter