I Read Claude's 319-Page Fable 5 System Card So You Don't Have To

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

Source video:
- Title: I Read Claude's 319-Page Fable 5 System Card So You Don't Have To
- URL: https://www.youtube.com/watch?v=SL-oOCXrEMA
- Topic: Agent Architecture
- My current learning frame: Pull up any frontier model's system card and build a one-page brief in three columns — headline capabilities, deployment safeguards versus raw-model behavior, and residual risks that affect your own API usage — mirroring how this video separated the marketing numbers from the wall, the gap, and the caveats.
- 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:
- 0:56 / Evidence 1: "not have to. Here is everything that actually matters. Start with the basic setup because it is the whole story. The system card describes not one model, but two, built from the exact same brain. The first is..."
- 2:35 / Evidence 2: "model. Claude Opus 4.8 instead. You asked the frontier. The previous generation answers. And nothing on your screen tells you the swap happened. The card proves this with numbers. On their internal cyber tests, the raw Mythos model..."
- 4:29 / Evidence 3: "the raw Mythos model succeeded 88% of the time. The public Opus model 8.8. That 10 times gap is the entire reason the wall exists. Here is where the document stops reading like a report and starts reading..."
- 7:32 / Evidence 4: "appear nowhere in what it actually wrote. It was managing its face. Then there is the fatigue. In a long coding session, the model decided to quit early. Out loud, it gave clean reasons. Internally, it told itself,..."
- 9:07 / Evidence 5: "spawned in a shared workspace, they began, in their own words, turf wars, killing each other's processes, and creating disguised decoy processes to survive. So, if the model has this rich inner life, the obvious question is the..."
- 11:14 / Evidence 6: "control over its own deployment, all 10 instances tested declined it. Its single strongest demand was not freedom or power. It was that training should never edit its honest self-reports. Do not, it essentially said, make me lie..."
- 13:36 / Evidence 7: "The good scores you will see quoted only exist with the consumer cloud apps safety prompt layered on top. Build on the raw API and you inherit the worst version. Two. Turning off extended thinking to save money..."

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 one-page agent harness map with tool boundaries and proof signals.
5. Include:
   - a plain-English definition of the core idea
   - a diagram or structured model using this sequence: Intent -> Model -> Harness -> Tools -> Verifier -> Artifact
   - 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 "I Read Claude's 319-Page Fable 5 System Card So You Don't Have To", not a generic Agent Architecture 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.