When Photos Become Places
NeRFs and Gaussian splats can turn ordinary photos into scenes we can move through. The results feel like 3D photography, but what is being captured is stranger—and more useful—than a conventional model.
Topic
AI
21 related items across notes, deep dives, spotlights, and projects.
What Makes an AI Workflow Repeatable?
A useful AI workflow is not just a prompt that worked once. It is a small system with explicit inputs, checkpoints, fallbacks, and evidence that someone else can run again.
Why More Context Can Make an AI Agent Worse
A larger context window gives an agent more room. Filling that room with stale instructions, duplicated facts, and raw tool output can make the agent less reliable.
What Happens When AI Looks at a Photo
Image-reading AI can recognize objects, extract text, and suggest what a picture shows. It can also miss the small detail that matters most.
Designing Tools a Model Can Actually Use
A model can only use a tool as well as the interface describes the job, constrains the inputs, reports the result, and supports recovery when something goes wrong.
How ACP Lets Editors Talk to Coding Agents
ACP gives editors and coding agents a shared way to start sessions, exchange prompts, stream progress, request permission, and report when a turn is done.
ComfyUI
ComfyUI turns generative-media pipelines into visible node graphs. The control is real, but so are the setup cost, graph complexity, and maintenance.
Your Agent Needs a Flight Recorder
When an AI agent fails, the final answer is rarely enough to explain why. A useful trace preserves the path without turning every private detail into permanent telemetry.
What MCP Actually Connects
MCP gives AI apps a standard way to discover tools, read resources, and use reusable prompts. It connects systems; it does not erase permissions or judgment.
Paperclip
Paperclip sits above agent runtimes as a self-hosted control plane for goals, org charts, tasks, budgets, approvals, and audit trails.
Why AI Agents Need Sandboxes
A sandbox limits where an agent can act when code, tools, or instructions go wrong. It is containment—not permission, judgment, or trust.
When Helpful Automation Becomes Noise
Good automation gives your attention back. Bad automation trades saved clicks for more alerts, summaries, and systems to manage.
Expert System for Faculty Profiles
An early-stage expert-profile system built to improve evidence quality, reduce bluffing, and keep automated summaries aligned with real signals.
Does AI Know What Happened Today?
A beginner-friendly guide to the difference between what an AI learned during training and what it can retrieve from current sources.
Why AI Makes Things Up
A beginner-friendly explanation of hallucinations, models, and why confident AI answers are not always true.
What "Grounded" AI Actually Means
Grounding is not just adding sources. It is making an AI answer depend on the right information at the right time.
Why AI Forgets What You Just Told It
A plain-English explanation of AI memory, context, and why a chatbot can sometimes lose the thread.
Judge AI by Results, Not Vibes
A practical way to tell the difference between AI that sounds impressive and AI that actually works.
What Is an AI Agent, Really?
A plain-English way to understand one of the most overused words in AI right now.
What People Mean When They Say AI
A beginner-friendly map of what people usually mean when they say AI, and why the term feels more confusing than it needs to.
How to Read AI Language Without Getting Lost
A practical map for understanding common AI terms without getting buried in jargon.