How a Knowledge Graph Made Haiku as Accurate as Fable 5
This video demonstrates, with a live side-by-side test, how injecting a knowledge graph via a deterministic prompt-submit hook lets the smallest model (Haiku) answer a tricky multi-hop question correctly and instantly, matching a much bigger model, while the same model without the graph gets completely confused.
Glitch Cat Club6 minTranscript found
Quick learning frame
Read this before watching.
A context/search lesson is about getting the right evidence into the agent at the right time through indexes, search, memory, or knowledge graphs.
New playlist item from Glitch Cat Club; queued for transcript-backed review, topic mapping, and a practical learning artifact.
Skill you build: The ability to use deterministic retrieval (a knowledge graph fired through a prompt-submit hook) to bypass a model's own reasoning and tool-calling for multi-hop factual lookups, so a small model performs like a large one.
Watch for the shift from claim to mechanism. The learning value is the point where the transcript reveals a repeatable action, tool boundary, context move, review habit, or artifact.
Concept diagram
Where this video fits.
01Work question
02Source inventory
03Index/search layer
04Retrieval rule
05Agent context
06Answer/proof
07Maintenance
Deep lesson
Turn this video into working knowledge.
1,073 cleaned transcript words reviewed across 320 timed caption segments.
Thesis
How a Knowledge Graph Made Haiku as Accurate as Fable 5 teaches a practical context/search move: This video demonstrates, with a live side-by-side test, how injecting a knowledge graph via a deterministic prompt-submit hook lets the smallest model (Haiku) answer a tricky multi-hop question correctly and instantly, matching a much bigger model, while the same model without the graph gets completely confused.
The goal is not to remember the video. The goal is to extract the operating principle, tie it to timestamped evidence, test how far the claim transfers, and make something reusable.
0:12
Knowledge graph closes the gap
“So, I listened. I've done what you said. I've created an artifact. I'm going to give you the link below. A knowledge graph made the smallest model, namely Haiku in this example, as accurate as the biggest. Haiku...”
A knowledge graph turned Haiku's wrong answer into the right one with zero tool calls and 2-millisecond retrieval, versus the ungrounded approach needing about 20 seconds of visible searching, proving the fix is faster and cheaper, not just more accurate. Write down one multi-hop question your own AI assistant regularly gets wrong (an answer that requires combining facts from two separate documents) as a candidate for a knowledge graph test.
1:19
How the hook works
“test case. You'll notice here I've got knowledge, the one with knowledge graph, and I've got the version with no graph, right? Let me give you the test case quickly so you understand it. Right, number one. Um...”
The mechanism is a deterministic prompt injection fired through a user-prompt-submit hook (available in Claude Code and OpenAI's tooling), meaning retrieval happens before the model reasons at all; it bypasses the model and tool calls entirely, so the model never has to search, it just receives the facts already assembled. Look up how your coding agent's prompt-submit hook works and sketch what a deterministic fact-injection step would look like for one of your own documents.
4:33
The multi-hop trap test
“my god, it it's just completely got confused. It's looking through my system, it doesn't have a clue. Okay, no problem. Let's try it with a knowledge graph. HiKu Use a prompt, submit memory, eight facts recalled in...”
Given the question 'a customer wants an $800 refund in March, who signs it off,' the answer required chaining three facts spread across separate documents (refunds over $500 need ops manager approval, the ops manager is Sarah, Sarah is out for March and Marcus covers); without the graph, Haiku got confused and couldn't even identify the right organization, but with the graph it recalled 8 facts in 2 seconds and answered 'Marcus Webb' correctly. Build a deliberately tricky multi-hop test case for your own use case, with the answer split across at least two documents and one decoy, and run it through your model with and without deterministic retrieval.
01
Work question
Start with this video's job: This video demonstrates, with a live side-by-side test, how injecting a knowledge graph via a deterministic prompt-submit hook lets the smallest model (Haiku) answer a tricky multi-hop question correctly and instantly, matching a much bigger model, while the same model without the graph gets completely confused. Treat "Work question" as the outcome you are trying to make visible, not a topic label. Anchor it to 0:12, where the video says: “So, I listened. I've done what you said. I've created an artifact. I'm going to give you the link below. A knowledge graph made the smallest model, namely Haiku in this example, as accurate as the biggest. Haiku...”
02
Source inventory
Use "Source inventory" to locate the part of the context/search mechanism the video is demonstrating. Ask what changes in your real setup if this claim is true. Anchor it to 1:19, where the video says: “test case. You'll notice here I've got knowledge, the one with knowledge graph, and I've got the version with no graph, right? Let me give you the test case quickly so you understand it. Right, number one. Um...”
03
Index/search layer
Turn "Index/search layer" into the reusable artifact for this lesson: A context retrieval map with source inventory, indexing/search path, query rules, freshness checks, and agent handoff. This is where watching becomes something you can inspect and reuse.
04
Retrieval rule
Use "Retrieval rule" as the application surface. Decide whether the idea touches a browser flow, a local file, a model choice, a source document, a UI, or a review step.
05
Agent context
Use "Agent context" to prove the lesson. The evidence should connect back to the video title, transcript anchors, and a concrete output, not a generic best-practice claim.
06
Answer/proof
Use "Answer/proof" to carry the idea forward: save the prompt, checklist, diagram, or operating rule that would make the next agent run better.
07
Maintenance
Connect "Maintenance" to How a Knowledge Graph Made Haiku as Accurate as Fable 5 by naming the claim, the evidence, and the artifact it should produce.
Example
Source-backed artifact packet
Convert the video into a scoped artifact request that includes the transcript claim, mechanism, acceptance criteria, and proof. The output should be a context retrieval map with source inventory, indexing/search path, query rules, freshness checks, and agent handoff..
Example
Context/search proof brief
Separate what the speaker claims, what the demo actually proves, and what still needs outside verification before you adopt the context/search pattern.
Example
Teach-back module
Transform the lesson into a definition, a Work question -> Source inventory -> Index/search layer -> Retrieval rule -> Agent context -> Answer/proof -> Maintenance diagram, one misconception, one practice exercise, and a check-for-understanding question.
Do not learn it wrong
Treating the title as the lesson without checking what the transcript actually says.
dumping all context
stale memory
retrieval with no proof trail
Letting the lesson drift into generic context-window advice.
Letting the lesson drift into memory hype without retrieval rules.
Letting the lesson drift into source claims without freshness checks.
Do not count this as learned until these are true.
01
State the transcript-backed claim in your own words: This video demonstrates, with a live side-by-side test, how injecting a knowledge graph via a deterministic prompt-submit hook lets the smallest model (Haiku) answer a tricky multi-hop question correctly and instantly, matching a much bigger model, while the same model without the graph gets completely confused.
02
Explain the practical stakes without hype: New playlist item from Glitch Cat Club; queued for transcript-backed review, topic mapping, and a practical learning artifact.
03
Map the idea onto the Work question -> Source inventory -> Index/search layer -> Retrieval rule -> Agent context -> Answer/proof -> Maintenance sequence and name the weakest link.
04
Produce the artifact and include the evidence that proves it: A context retrieval map with source inventory, indexing/search path, query rules, freshness checks, and agent handoff.
Put it into practice
Give this grounded prompt to Codex or Claude after watching.
You are helping me turn one specific YouTube video into real, durable learning.
Source video:
- Title: How a Knowledge Graph Made Haiku as Accurate as Fable 5
- URL: https://www.youtube.com/watch?v=svvlT-nX6N8
- Topic: Codex + Claude Workflows
- My current learning frame: Clone the linked GitHub repo, follow the getting-started guide to set up a user-prompt-submit hook with your own small ontology, and run the same model twice (with and without the graph) on a self-authored multi-hop trap question to see the accuracy gap firsthand.
- Why this matters: New playlist item from Glitch Cat Club; queued for transcript-backed review, topic mapping, and a practical learning artifact.
Transcript anchors from this exact video:
- 0:12 / Evidence 1: "So, I listened. I've done what you said. I've created an artifact. I'm going to give you the link below. A knowledge graph made the smallest model, namely Haiku in this example, as accurate as the biggest. Haiku..."
- 1:19 / Evidence 2: "test case. You'll notice here I've got knowledge, the one with knowledge graph, and I've got the version with no graph, right? Let me give you the test case quickly so you understand it. Right, number one. Um..."
- 3:02 / Evidence 3: "it explains why. Wonderful. Okay. Now, let's go to the no uh the graph example. Same thing. Fable 5 on XHI auto mode. We're going to give it the same exact prompt. Customer wants $800 refund. Let's go."
- 4:33 / Evidence 4: "my god, it it's just completely got confused. It's looking through my system, it doesn't have a clue. Okay, no problem. Let's try it with a knowledge graph. HiKu Use a prompt, submit memory, eight facts recalled in..."
Video-aware target:
- Prompt lane: Context/search
- Mechanism to extract: Extract how context is found, filtered, refreshed, and handed to the agent before it acts.
- Artifact to produce: A context retrieval map with source inventory, indexing/search path, query rules, freshness checks, and agent handoff.
- Artifact must include: source inventory; index/search layer; query rule; freshness check; agent handoff; proof behavior
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, transcript support, what the demo proves, confidence, and what still needs verification.
3. Extract the actual teachable mechanism from the video: Extract how context is found, filtered, refreshed, and handed to the agent before it acts. Do not invent claims that are not supported by the title, lesson frame, or transcript anchors.
4. Build a reusable learning artifact: A context retrieval map with source inventory, indexing/search path, query rules, freshness checks, and agent handoff.
5. Include:
- a plain-English definition of the core idea
- a diagram or structured model using this sequence: Work question -> Source inventory -> Index/search layer -> Retrieval rule -> Agent context -> Answer/proof -> Maintenance
- answers to these source questions: What source is searched or indexed? | What query/retrieval rule is demonstrated? | How does the agent use the retrieved context?
- 3 concrete examples that apply the video idea to real agentic work, such as codebase memory; personal wiki retrieval; Elastic search context engineering
- 2 failure modes the video helps prevent, chosen from the transcript evidence and these likely risks: dumping all context; stale memory; retrieval with no proof trail
- a checklist for the next real workflow, focused on: sources, query, freshness, handoff, citation/proof
- one practical exercise with a clear done signal: Write three retrieval queries for one real project and define what each must return.
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 "How a Knowledge Graph Made Haiku as Accurate as Fable 5", not a generic Codex + Claude Workflows essay.
- Cite the transcript wherever the prompt names a task boundary, review habit, context move, or verification standard.
- Prefer operational examples, failure modes, and reusable artifacts over broad definitions.
- Call out uncertainty instead of smoothing over weak evidence.
- Avoid these generic drifts: generic context-window advice; memory hype without retrieval rules; source claims without freshness checks.
- If evidence is weak or missing, stop and say what transcript segment or timestamp needs review instead of guessing.
- Finish with a concise artifact I could paste into my learning app.
Misconceptions
What to stop believing.
One agent should do every task.
Different tools have different strengths. Routing is part of the workflow.
More context is always better.
Relevant context helps; stale context causes drift and cost.
Practice studio
Learning only counts when you make something.
01
Transcript evidence map
Separate what the video actually says from what you already believe about the topic.
3 source-backed takeaways with timestamps, confidence, and a transfer note.02
One useful artifact
Apply the video to a real workflow and produce a context retrieval map with source inventory, indexing/search path, query rules, freshness checks, and agent handoff..
A reusable artifact with a done signal and one verification step.03
Context/search teach-back card
Explain the context/search mechanism to someone who has not watched the video yet.
A 90-second explanation, one diagram, one example, and one misconception to avoid.
Recall check
Answer first, then reveal — without rewatching.
What two measurable improvements did the knowledge graph give Haiku on the test question, according to the presenter?
What mechanism delivers the knowledge graph facts to the model, and why does that mean it bypasses the model entirely?
In the multi-hop trap test, what three chained facts did the model need to combine to correctly answer who signs off an $800 refund, and what happened to Haiku without the graph?
Source shelf
Use the video as a doorway, then verify with primary sources.
