Total runtime: 20 minutes
Deck: http://localhost:3001/slides/index.html (with npm run dev -- -H 0.0.0.0 -p 3001)
Navigation: → / ←, swipe, or on-screen controls · F fullscreen
Assets live in public/slides/. Drop the demo recording at public/slides/demo.mp4.
| Block | Minutes | Slides |
|---|---|---|
| Opening + demo | 0:00–4:00 | 1–2 |
| What you saw + problem | 4:00–7:00 | 3–5 |
| Agent + Nimble role | 7:00–10:00 | 6–7 |
| Feature breakdown | 10:00–14:00 | 8–12 |
| Differentiation + win | 14:00–17:00 | 13–14 |
| Build path + generalization + close | 17:00–20:00 | 15–17 |
Say:
“Morning espresso with Nimble. The thesis is not ‘AI for GTM.’ It is this: Nimble turns live web chaos into structured, source-grounded, agent-ready intelligence — before the LLM reasons.”
So what: Plant the differentiation before the demo so the video is evidence, not entertainment.
Do: Play the Morning Signal recording without narrating over every click.
Teach through the demo:
Say after video (10s): “You did not watch a chatbot. You watched a workflow with Nimble in the critical path.”
Walk the five boxes left to right. Pause on the two yellow/green Nimble steps.
Say: “Context and planning are necessary. Synthesis is necessary. But without Search and Extract, the model is writing fiction with confidence.”
Slide 4: Point at the constellation — launches, pricing, docs, community, jobs, analyst notes.
Say: “GTM does not have an information shortage. It has a freshness and assembly problem.”
Slide 5: Hit CRM, battlecards, LLM memory as three blind spots.
Say: “None of these know what changed on the public web this morning. That is the gap Nimble fills.”
Trace Context → Plan → Critical path → Output.
Emphasize: “The LLM is the dashed node at the end. LangGraph makes each step explicit. Nimble Search and Extract are highlighted because that is where live truth enters the system.”
Optional: Mention LangSmith traces as proof the graph is real.
Four blocks: discover → structure → audit → ground. Then the “Only then → synthesis” bar.
Say: “If you remember one diagram, remember this: synthesis is not allowed to run until evidence exists.”
Keep visible text minimal; speak the “why.”
| Slide | Punch line |
|---|---|
| Search | Finds fresh signals with controls developers actually use |
| Extract | Pages become citable evidence, not pasted HTML |
| Agents | Encode research playbooks; run them like infrastructure |
| Crawl / Map | Same pattern, site-scale when Morning Signal graduates from demo |
| Drivers | Standard / render JS / stealth = cost–control tradeoff |
Bridge: “Morning Signal uses Search + Extract today. Agents, Crawl, Map, and drivers are how you productionize the same pattern.”
Point to Exa/Tavily (retrieval), Firecrawl (extract/crawl), Apify (marketplace), Bright Data (heavy infra), then Nimble in the top-right.
Say: “The wedge is not ‘we also scrape.’ It is live search + structured extract + agents + crawl/map + pricing controls as one agent-ready platform.”
Scan the yellow column. Do not read every cell.
Say: “For this use case — source-grounded GTM intelligence — freshness, structure, inspectability, and agent fit have to land together. That is why Nimble wins Morning Signal.”
Read the six steps as a recipe: key → env → Search → Extract → LangGraph → synthesize with sources.
Say: “This is the DevRel standard: leave the room with a path they can run tonight.”
Pattern: Live web signal → structured evidence → grounded synthesis → business action.
Gesture across use cases: GTM, recruiting, market research, financial monitoring, product/competitive.
Say: “Morning Signal is one instance of a reusable cookbook pattern.”
Restate thesis. Offer Q&A on architecture, LangSmith, and the build path.
Say: “Prove the workflow. Don’t pitch the API. Thank you — questions.”
npm run dev -- -p 3001) for up to 4 minutes; keep DEMO_MODE ready.