Evaluation & AI Reliability · core
Prompt & Version Logging
Treating prompts as versioned artifacts with logged inputs/outputs.
Mental model
Treat prompts like code — every production change gets a version, a saved diff, and a re-run of your evals. Without that, when quality drops you have no commit to point at.
How to study Prompt & Version Logging
Begin by restating the mental model in your own words, then connect it to a concrete system you have built or operated. Name the mechanism, the constraint it addresses, and the trade-off it introduces. Use PostHog — LLM analytics to check details, but close the source before writing your explanation. Retrieval is the learning step; rereading is only preparation.
Next, compare Prompt & Version Logging with LLM Evals, Model Routing. Ask what changes in correctness, latency, resource use, operability, and failure recovery. Complete Diff two prompt versions and preserve the command, input, output, and one failed attempt as evidence. Finish by explaining the idea without jargon to someone who has not studied the track.
Proof of understanding
- Explain the mechanism from first principles and identify the state it reads or changes.
- Give one situation where the concept is the right choice and one where it is not.
- Predict a realistic failure mode before running the drill, then compare the prediction with evidence.
- Connect the result to a roadmap or build artifact instead of treating the concept as isolated trivia.
Learn from primary sources
- PostHog — LLM analytics (doc)
Practice and explain it back
Diff two prompt versions
v1 and v2 prompts differ by one instruction. Name what you must re-run before shipping v2.
Expected evidence: Frozen eval set on both versions; compare pass rate delta.
Open the interactive drill →Review prompts
- Quality dropped last Tuesday. What must have been logged for you to find the cause?
Build evidence
Synthesize: Evaluation & AI Reliability
Build an evidence-backed evaluation and observability system for models, tools, and agents. Produce one working system, benchmark, or evidence-backed design that integrates the path.
- Implements or precisely models the core mechanisms from all three milestones
- Includes at least one injected failure or adversarial case and demonstrates recovery
- Reports quality, latency, resource, reliability, or usability measurements relevant to the domain
- Ships a concise architecture note explaining decisions, trade-offs, and remaining risks
Prerequisites
None assigned yet.