RAG Was Blind This Whole Time — Berkeley’s PixelRAG Gives Your AI Eyes

You are helping me turn one specific YouTube video into real, durable learning.

Source video:
- Title: RAG Was Blind This Whole Time — Berkeley’s PixelRAG Gives Your AI Eyes
- URL: https://www.youtube.com/watch?v=ih5GKdGUIns
- Topic: Creative Automation
- My current learning frame: Pick one chart-and-table-heavy document set you own, stand up a local PixelRAG index over it, and ask five questions whose answers live in visual structure — then decide per the video's rule whether each of your retrieval use cases needs visual or text search.
- Why this matters: New playlist item from Hyperautomation Labs; queued for transcript-backed review, topic mapping, and a practical learning artifact.

Transcript anchors from this exact video:
- 0:00 / Evidence 1: "Your AI has been reading documents with its eyes closed. Look at this table. Clean rows, clean columns. The answer sitting right there. But the moment a normal rag system touches it, this happens. It gets shredded into..."
- 1:55 / Evidence 2: "beautifully simple question. What if we never pass the page to text at all? What if we just take a picture of it? That is Pixel Rag. Instead of shredding the page, it screenshots it. It slices that..."
- 4:10 / Evidence 3: "vision model and becomes a vector. Four, build index. All those vectors go into a vector index called FAISS, the same battle-tested library that powers huge text search systems. And five, serve. A simple API takes your question..."
- 7:08 / Evidence 4: "going to love. PixelRag ships as a Claude code plugin called PixelBrowse. Install it and Claude can screenshot any page and actually see it. Charts, tables, dashboards, the way a human does, instead of choking on scraped text."
- 8:55 / Evidence 5: "compute, and retrieval is a little slower. On plain prose, an article that is just paragraphs with no real visual structure, the gains are small. You might not need it there. And the full pre-built Wikipedia index is..."
- 10:32 / Evidence 6: "If you want to go deeper, my complete Claude code guide is linked down in the description. Rag was blind this whole time. Now, you can finally see the fix. Go build with it."

Your task:
1. Use the transcript anchors above as the primary source packet. If you add outside context, label it clearly as outside context and keep it secondary.
2. Create a source-check table with columns: timestamp, claim, what the demo proves, confidence, and what still needs verification.
3. Extract the actual teachable claims from the video. Do not invent claims that are not supported by the title, lesson frame, or transcript anchors.
4. Build a reusable learning artifact: A creative workflow board with critique criteria and review checkpoints.
5. Include:
   - a plain-English definition of the core idea
   - a diagram or structured model using this sequence: Brief -> Source -> Generation -> Selection -> Edit -> Taste Review
   - 3 concrete examples that apply the video idea to real agentic work
   - 2 failure modes the video helps prevent
   - a checklist I can use the next time I run Codex or Claude
   - one practical exercise with a clear done signal
6. Add a "learning transfer" section: what changes in my workflow tomorrow if I actually learned this?
7. Add a "source check" section that cites which transcript anchor supports each major takeaway.

Quality bar:
- Make this specific to "RAG Was Blind This Whole Time — Berkeley’s PixelRAG Gives Your AI Eyes", not a generic Creative Automation essay.
- Prefer operational examples, failure modes, and reusable artifacts over broad definitions.
- Call out uncertainty instead of smoothing over weak evidence.
- If evidence is weak, say what transcript segment or timestamp needs review instead of guessing.
- Finish with a concise artifact I could paste into my learning app.