OpenHuman: Local-First AI Agent That Remembers Everything About You

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

Source video:
- Title: OpenHuman: Local-First AI Agent That Remembers Everything About You
- URL: https://www.youtube.com/watch?v=Whxv-D51ZPE
- Topic: Interfaces + Open Design
- My current learning frame: Diagram OpenHuman's ingestion pipeline from a connected Gmail label to a topic tree, annotating where chunking, scoring, hotness, token-juice compression, and the 20-minute AutoFetch each act.
- Why this matters: New playlist item from AI Stack Engineer; queued for transcript-backed review, topic mapping, and a practical learning artifact.

Transcript anchors from this exact video:
- 0:00 / Evidence 1: "Open Human is something every AI agent quietly fails at. You explain your project to the thing, your tools, how you work, and then you close the window, and it forgets all of it. Next session, you're back..."
- 1:34 / Evidence 2: "the memory system. The point they keep hammering is that every model out there, all 200-plus of them, is stateless. You send a prompt, you get an answer, and the context just evaporates. Even the ones with so-called..."
- 3:12 / Evidence 3: "a quick score, happens immediately, so the app never freezes. The heavy lifting, the embeddings, and the entity extraction runs in the background. And if you turn on local AI through a llama, even that heavy work can..."
- 4:53 / Evidence 4: "70% cut on his own data, a bit under the 80% the company claims, but still a big difference when an uncompressed sweep through a top model can run you 20 or 30 bucks. On top of that,..."
- 7:27 / Evidence 5: "local-first design backs that up. Your raw data doesn't leave your machine unless you specifically put it into a prompt. But you're concentrating a lot of sensitive access in one place, email, code, calendar, all at once. So,..."
- 8:58 / Evidence 6: "leave it? What Open Human gets genuinely right is treating memory as the actual product, not a feature stapled on the side. Most agents add memory as an afterthought. This one is built around the memory tree, and..."

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 UI critique sheet for judging whether an AI interface improves control.
5. Include:
   - a plain-English definition of the core idea
   - a diagram or structured model using this sequence: Intent -> Canvas -> Artifact -> Preview -> Feedback -> Iteration
   - 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 "OpenHuman: Local-First AI Agent That Remembers Everything About You", not a generic Interfaces + Open Design 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.