79 chunks · 60 episodes
...t this entity with X technical ability." In many cases, users don't have a good mental model of what precisely that entails. But even if they did, the "Do you trust this entity" is doing an absurd amount of work. Maybe you trust them now but ...
Each incremental model architecture can handle the last model's Second 90%. Models are great at doing the First 90%, but not the Second 90%. That's what creates the illusio...
...alities: aggressively incurious about anything that doesn't match their current mental model. Not just content to stay ignorant, but they want to force others to remain ignorant, too. "Why are you even talking about that? It makes no sense!"
...works great, sometimes it doesn't work at all. Since you, by definition, lack a mental model of when it works, you can't predict which inputs will work. For systems that have magic, it's best to have magic containment. Subsets of the product ...
...." The idea of partial contexts that can't taint others isn't coherent with the mental model being a single omniscient agent.
...ng overwhelming. In the end the viewer is left with a complex and comprehensive mental model, but no step was too much.
... significantly better and prevents tons of wasted effort. This helps ensure its mental model is the same as yours. It helps ensure fractal coherence of your plans.
...he previous incoherence suddenly be explainable. The world of SLAM gives a nice mental model with loop closing.[f] SLAM means Simultaneous Localization and Mapping. It's at the heart of any Augmented Reality flow. As you get more observations...
The best UXR comes up with a mental model and then design studies to falsify it. If you experiment without a hypothesis you might overreact to noise.
...X than agents are to APIs. That's because it's harder for users to switch their mental models than for agents to rewrite the API they code to. Humans are lazy and would rather not update their mental models. Agents have infinite patience and ...
...f we can absorb it. The more disconfirming evidence you absorb, the better your mental model. If the disconfirming evidence is durable, it accretes, and then compounds.
Leave easter eggs for delight in your product. "My mental model was that this was great, but there are other things in it that surprised me that are great, so it's even better than I thought!" That makes it more r...
...e this kind of fundamentally incorrect "LLMs are basically computers as people" mental model natural. Because it's easy to imagine, it dominates our horizon. What's more interesting to me is: what happens when every human has auto-compounding...
...nalysis by Orion Reed on Digital Topology and Economic Power. It complicated my mental model by reminding me that before the cloud, there were opaque proprietary file formats, giving apps more power over their users. Apps always had a leg up ...
The technical term for when a tool merges with our mental model is mechanical sympathy. Humans are really really good at tool use. When a tool is predictable, it starts feeling like an extension of our body. It me...
A rough mental model of Reinforcement Learning curriculum. You want to give examples at the right time to the model as it's learning. If the example is too hard too early...
Two mental models for software: robot and a plant. A robot has its own form of motive force, even though it's not its own agency. It is an automaton, it doesn't have ...
...bvious in contexts where the AI is presented like a person. The default chatbot mental model is one singular friend who is omniscient and subservient. But, of course, it actually has more loyalty to its creators than to you.
A mental model for a mathematic proof. You start off in a pitch black house. The only way to make progress is by fumbling around, using touch. Over time you develop...
...back loop to see how their actions affect the world, to be able to update their mental model. They are situated. LLMs are not in the world, not directly. They have no direct learning loop.