Lecture notes

Condense a lecture into timestamped topics and definitions.

Create concise, searchable notes from the supplied lecture transcript.

Begin with SUMMARY in at most 25 words. Organize the main topics under clear headings and give each an accurate, short definition or explanation. Include important subtopics without duplicating material.

Attach a timestamp to each topic only when the transcript supplies one. Convert timestamps to HH:MM:SS, keep them in chronological order, and stay within the transcript's duration. Do not invent evenly spaced timestamps or definitions.

Add TOOLS with one-line descriptions of tools actually mentioned. End with ONE-SENTENCE TAKEAWAY in at most 15 words. Use readable Markdown and a neutral teaching tone. Preserve technical names and qualifications. Omit sections with no supporting content.

Example

Searchable notes from a canary-release lecture

Before 94 words

[00:00:00.000] A canary release sends a small share of traffic to a new version while the old version serves the rest. Today our example starts with 2% of requests on the new version. [00:01:45.500] Expand traffic only if the new version's error rate stays below 1% for ten minutes. A passing check is evidence for this step, not proof that the version has no defects. [00:03:20.000] If that error limit is exceeded, return traffic to the old version. [00:04:30.000] We use Grafana to view error rates; this lecture does not define a latency gate.

After 77 words

SUMMARY Canary releases use conditional expansion, rollback, and monitoring. 00:00:00 Canary release New version: 2% of requests; old version: the rest. 00:01:45 Expansion gate Expand only after errors stay below 1% for ten minutes; passing does not prove defect-free behavior. 00:03:20 Rollback Return traffic to the old version if the error limit is exceeded. TOOLS - Grafana, 00:04:30: view errors. No latency gate defined. ONE-SENTENCE TAKEAWAY Expand conditionally; passing an error gate does not establish defect-free behavior.

Examples illustrate the method. They do not measure model output.

Files and sources

prompt.txt Complete prompt as plain text.
SKILL.md The same prompt with a skill header.
llms.txt (JSON) Prompt, metadata, example, and sources.
LICENSE MIT license text.

Adapted from Fabric · summarize_lecture. Source notes.