Turn a Long RFC into an Exec Summary with an LLM

Long docs fail when you dump without a job

Pasting a 5k-word RFC and saying “summarize” wastes both attention and credits. Tell the model the output shape: executive bullets, outline, risks, and this-week actions for a specific role (eng lead, PM, SRE). The playground’s “Long doc → summary” demo uses a credit-metering RFC as filler so you can see the shape before swapping in your own text.

Structure beats raw length

• Keep headings and numbered sections — they become citation anchors.
• Put the ask above the document body so instructions are not buried.
• Trim appendices and duplicate changelogs; they inflate the live token estimate with little signal.
• For multi-file packs, lead with a table of contents, then the 3–10 hottest files.

Spend control on spacebunny.me

This site hard-caps around 100k tokens per request. Free accounts should treat that as a guardrail, not a goal. Use the composer’s next-send credit estimate (input + a rough reply budget). If you are near zero, top up with the $10 pack while Space Bunny Alpha is still preview — monthly plans make less sense when upstream availability wobbles.

Validate the summary

Ask the model to tag each risk or action with a section header from your paste. If citations drift, the model is improvising — shorten the doc or raise temperature down and re-ask for grounded bullets only. Export Markdown when the exec summary is good enough to share outside the chat.