reorganize intro slides: improve structure, remove double headers, drop OpenAI Dev Day reference and 'What's Changed' section

Author: jxnl

intro-slides/slides.md

@@ -20,7 +20,7 @@ slideNumber: true
 
 # Systematically Improving RAG Applications
 
-September 16, 2025, `maven.com/applied-llms/rag-playbook`
+November 19, 2025
 
 <br>
 
@@ -36,18 +36,16 @@ By Jason Liu
 
 ---
 
-# Today’s Plan
+# Today's Plan
 
 - Introductions
 - About me; consulting and training
-- Who are you?
-- Format & office hours
+- Course format & logistics
 - Key insights & course outcomes
-- What’s changed since v1 (2024)
 - Syllabus (Sessions 0–3)
 - Syllabus (Sessions 4–6)
-- Resources & contributions
-- Q&A and next steps
+- Resources & next steps
+- Q&A
 
 ---
 
@@ -63,7 +61,7 @@ By Jason Liu
   - CV + multimodal retrieval
   - VAEs/GANs for GenAI
   - ~$50M incremental revenue
-  - led ~$400K/yr data curation for next‑gen models
+  - Led ~$400K/yr data curation for next‑gen models
 
 ---
 
@@ -72,40 +70,53 @@ By Jason Liu
 - Personal note: Hand injury (2021–2022) → shifted focus to higher‑leverage teaching and advising
 - Consulting (2023–present): Query understanding, prompts, embedding search, fine‑tuning, MLOps/observability; upgrading legacy workflows to agentic systems
 - Clients: HubSpot, Zapier, Limitless, and others across assistants, construction, research
+- Recently: helping with medical triage AI; advising startups that do observability
 
 ---
 
-# Format & Office Hours
+# Course Format
 
-- Inverted classroom: ~6 hours pre‑recorded lectures + tutorial videos + Jupyter exercises
-- Slack: post questions in the cohort channel for async help
-- Welcome to share your learnings online via linked in or twitter (please link back to us some how)
+**This is a 3-week accelerated version of the course.**
 
-<v-click>
+- Inverted classroom: ~5 hours pre‑recorded lectures + tutorial videos + Jupyter exercises
+- I recommend watching two sessions per week
 
-## Office hours:
+---
+
+# Office Hours & Community
+
+- **Office hours**: Bring your problems, introduce yourself
+  - Treat it like a tech‑lead review of your work
+  - Cameras on is really appreciated! Helps me a lot.
+  - Guest lectures: 1-2 times a week, practitioners actively building in the space
+- **Slack**: For any questions about code or anything else, please post on the Slack
+- **Sharing**: Welcome to share your learnings online via LinkedIn or Twitter (please link back to us somehow)
 
-- bring your problems, introduce yourself.
-- treat it like a tech‑lead review of your work
-- cameras on is really appreciated! helps me a lot.
-- Guest lectures: 1-2 times a week, practitioners actively building in the space.
+---
 
-</v-click>
+# Logistics & Support
 
-<v-click>
+- **Scheduling**: Occasional reschedules; advance notice
+- **Credits/support**: Contact Marian — [email protected]
+- **Fit concerns**: If you feel like it's not a really good fit for you, just message me and we can figure out how we can make this better for you
 
-## Logistics
+---
 
-- Scheduling: occasional reschedules (e.g., OpenAI Dev Day); advance notice
-- Credits/support: contact Marian — support at jxnl.co ([email protected])
+# Upcoming Talks This Season
 
-</v-click>
+- Understanding tool called hallucinations: how LangChain themselves have been thinking about building agents
+- Dropbox talk: how they think about knowledge graphs, DSPy, and a bunch of new topics
+- Stay tuned to Slack to figure out when these events will happen
 
 ---
 
-# Key Insights & Course Outcomes
+# What You'll Learn
 
-This course will give you the foundations and practical skills to build, evaluate, and operate retrieval-augmented generation (RAG) systems. Here’s what to keep in mind and what you’ll learn:
+This course will give you the foundations and practical skills to build, evaluate, and operate retrieval-augmented generation (RAG) systems.
+
+---
+
+# Core Principles
 
 ### Keep these in mind
 
@@ -120,104 +131,76 @@ This course will give you the foundations and practical skills to build, evaluat
 
 # The Models are Good but Context Is the Bottleneck
 
-<v-click>
-
 - Models are already very capable for work.
 - Even if models hold steady, apps can still improve.
 - With the right context, success rates are very high.
 - The real challenge is getting that context — the R in retrieval.
 
-</v-click>
-
----
-
-# What’s Changed Since v1 (2024)
-
-Lets talk more about this in the office hours!
-
-- Since Claude 3.5, tools/agents are much more reliable; planning loops keep improving
-- Code assistants (e.g., Claude Code) excel at code RAG flows: grep + edit with validation
-- Prompt caching widespread; longer contexts change cost/latency tradeoffs
-- Better PDF/diagram extraction; richer citations in UIs (bounding boxes)
-- Shift from one‑off evals to continuous testing and monitoring over time
-- Always a tradeoff: performance × latency × cost; be mindful of pricing impacts
-- “Context engineering” is emerging; 2026 update of this course will focus more here
-
 ---
 
-# Syllabus (Sessions 0–3)
+# Sessions 0–3: Foundations
 
-- Session 0: Product mindset; RAG as a recommender; improvement flywheel
-- Session 1: Synthetic data and retrieval evals; precision/recall; baselines
-- Session 2: From evals to training data; reranking; embedding fine-tuning
-- Session 3: UX that collects data; streaming; chain of thought; validation
+- **Session 0**: Product mindset; RAG as a recommender; improvement flywheel
+- **Session 1**: Synthetic data and retrieval evals; precision/recall; baselines
+- **Session 2**: From evals to training data; reranking; embedding fine-tuning
+- **Session 3**: UX that collects data; streaming; chain of thought; validation
   - Pacing note: Week 3 is intentionally lighter—use it to catch up and get ahead
   - Focus on UX patterns; not the most critical week content‑wise
 
-<v-click>
-
-## Main takeaway
+---
 
-- fast retrieval evals (precision/recall on key chunks)
-- rerank/fine‑tune to get a 10-20% improvement
-- deploy and collect real data via UX
+# Sessions 0–3: Main Takeaway
 
-</v-click>
+- Fast retrieval evals (precision/recall on key chunks)
+- Rerank/fine‑tune to get a 10-20% improvement
+- Deploy and collect real data via UX
 
 ---
 
-# Syllabus (Sessions 4–6)
+# Sessions 4–6: Advanced Topics
 
-- Session 4: Topic modeling; query segmentation; prioritization frameworks
-- Session 5: Specialized indices; multimodal search (docs, images, tables, SQL)
-- Session 6: Query routing; tools-as-APIs; single vs multi-agent; measurement
+- **Session 4**: Topic modeling; query segmentation; prioritization frameworks
+- **Session 5**: Specialized indices; multimodal search (docs, images, tables, SQL)
+- **Session 6**: Query routing; tools-as-APIs; single vs multi-agent; measurement
   - Week 6 is lighter; focus on routing and preview the context‑engineering direction
 
-<v-click>
-
-## Main takeaway
+---
 
-- Figure out whats important to you and your users
-- Build specialized indices for those usecases
-- Make sure the Agent is able to use the specialized indices
+# Sessions 4–6: Main Takeaway
 
-</v-click>
+- Figure out what's important to you and your users
+- Build specialized indices for those use cases
+- Make sure the agent is able to use the specialized indices
 
 ---
 
-# Resources & Contributions
-
-Feel free to share this with coworkers, but don't post these links on social media. You can completely welcome to write your own notes and share them online! (Please link back to us some how)
+# Resources
 
-<v-click>
-
-## Resources
+Feel free to share this with coworkers, but don't post these links on social media. You're completely welcome to write your own notes and share them online! (Please link back to us somehow)
 
 ![QR Codes for Resources](./assets/images/codes.jpeg)
 
 - Study notes (work in progress): https://567-labs.github.io/systematically-improving-rag/
 - Talks/"greatest hits": https://567-labs.github.io/systematically-improving-rag/talks/
 - Slack: https://join.slack.com/t/improvingrag/shared_invite/zt-3dkinqb3q-vknvaBLoTx5tBj4PpGOVjw
 - Contribute via PRs/issues; add examples; suggest edits
-</v-click>
 
-<v-click>
+---
 
-## Recommendations
+# Recommended Talks
 
-- Skylar’s RAG anti‑patterns Talk
+- Skylar's RAG anti‑patterns Talk
 - Anton's Text Chunking Strategies Talk
 - Exa's Why Google Search Sucks for AI Talk
 - Colin's Agentic RAG Talk
-</v-click>
 
 ---
 
-# Q&A
+# Q&A and Next Steps
 
 Short Q&A (About the Class Format) and then we'll let you watch the first videos
 
-1) Find 'Optional: Watch Lecture' on your Calendar as a shortcut
-2) Feel free to watch at 2x speed!
+1. Find 'Optional: Watch Lecture' on your Calendar as a shortcut
+2. Feel free to watch at 2x speed!
 
-See you at the office hours!
+See you at the office hours!