#47 | See it, fix it, design it

TL;DR: Moving house taught me how AI actually solves real problems (versus the marketing hype).

👋 Hello,

Let me apologize for missing last Friday’s edition. Grrh!

Moving house with a 12-week-old while battling a cold isn’t ideal timing. So, I had to break my streak. Damn it. Still, the move prompted me to reflect on how I actually utilize AI when life demands practical solutions.

I’ve been systematically building my approach—e.g., organizing Perplexity Pro spaces, creating visual problem-solving workflows, and getting more structured about AI assistance.

The system suits me. Ok, I keep wondering if there are smarter ways to do this (probably are).

With AI evolving quickly, I figure I’m trying to keep pace like everyone else. And because my house move is and was the overall dominating topic in the last two weeks, I can at least use this topic to explore what’s actually possible versus what’s mostly hype.

When corners need solutions

Everyone is likely familiar with this. You stand in your new place, staring at an awkward corner space. Damn, it’s too small for a bookshelf, too big to ignore.

My old approach? Stare at it for weeks, maybe ask friends, probably end up with something that doesn’t work.

My current approach? Pull out phone, snap photo, upload to organized Perplexity space, get three specific solutions in two minutes.

The process: Photo → AI analysis → solution → implementation.

Simple? Yes. Effective? Definitely. Consistency matters when you’re dealing with spatial challenges at 9 PM (believe me on this one).

Although I often wonder what more systematic approaches might look like…

My visual AI setup

Here’s the framework I’ve developed during the move: systematic visual workflows for home challenges.

The basic flow: Take photo → upload to designated Perplexity space → receive structured solution → implement

My Perplexity Pro organization:

• “Home systems oracle”: Product manuals, warranty information, maintenance history
• “Visual space designer”: Room photos, style preferences, budget constraints
• “Project coordinator”: Active projects, timelines, contractor information

Each space handles specific problem types. The system’s oracle tells me which drill bit works for different walls (finally got that right).

The visual designer suggests furniture arrangements that make sense. The project coordinator keeps everything organized.

Multi-modal AI—that’s AI processing different input types like text, images, and voice simultaneously—makes this possible.

Think of it as having an assistant that can see what you see, understand what you’re describing, and connect both for actionable advice.

This approach builds knowledge bases that get smarter with each question (beats starting from scratch every time).

Though I suspect there are much more sophisticated setups possible…

Real applications that work

Let me show you what systematic visual AI looks like in practice:

Problem-solving uses:

• Wall damage assessment: Upload photo → get repair method based on wall type and damage severity
• Electrical troubleshooting: Show outlet placement → receive safety assessment and guidance
• Plumbing mysteries: Photo weird connections → understand exactly what parts you need

Design applications:

• Room analysis: Photos + measurements → furniture arrangements with traffic flow considerations
• Lighting strategy: Natural light assessment → artificial lighting recommendations
• Style consistency: Visual check → AI identifies what fits your aesthetic

Manual-free operations:

• Router configuration: Photo current settings → get security setup for your needs
• Appliance optimization: AI analyzes usage patterns → suggests ideal settings
• Smart device integration: Visual troubleshooting when connections fail

The idea lies in having systems that remember what works for your specific space, style, and constraints.

I’m discovering this systematic approach works well, though I suspect some might be thinking about more automated solutions… (or think it is BS nonsense overall).

Learning what’s actually possible

While this could be on the cards for some people, it is not necessarily what I want, choose, and actually do, ok.

However, I’m interested in understanding what’s possible and what the mainstream market may offer in the future.

For example, consider proactive home intelligence instead of reactive photo-solving. AI that monitors, predicts, and coordinates before problems surface.

Multi-agent systems:
Specialized AI assistants for different home functions that communicate with each other, learn your patterns, and handle routine decisions.

These use persistent memory, retaining context from previous interactions rather than starting fresh each time.

Though based on current technology, most agent systems still require significant manual setup and “lose all context once a workflow ends,” which limits their autonomy.

Multi-modal integration:
Combining visual analysis with sensor data and smart home information. IoT devices—sensors, smart thermostats, security cameras—feed continuous data to AI systems for a comprehensive understanding.

Home Assistant recently launched AI task integration that “allows you to attach files or cameras and ask it what is happening,” though it requires manual configuration and monitoring.

Predictive intelligence:
AI that identifies maintenance issues before they become expensive problems, automatically orders supplies based on project progress, and adjusts home systems seasonally.

The learning curve and investment feel steep, though (and the reality often doesn’t match the promise).

The evolution path

So, in theory, what began as simple problem-solving during a move revealed something larger.

We’re approaching AI systems that prevent problems rather than solving them after they occur.

Three levels of visual AI use:

Level 1:
Reactive Photo → solution (where most people start)

Level 2:
Systematic Organized workflows that accumulate knowledge over time

Level 3:
Proactive AI agents that monitor, predict, and act with minimal intervention

Most people stay at level one. I consider myself level two and eyeing level three. Level three? That’s where things get intriguing (though also more complex than advertised).

For those already experimenting with more automated approaches…

The technical reality

If systematic Perplexity spaces feel basic for your needs, here’s what’s actually being built:

Multi-agent orchestration
Specialized agents for different home functions—maintenance, design, procurement—with decision-making hierarchies. Though IBM experts note that “2025 might be the year we go from experiments to large-scale adoption,” while warning about the need for “strong compliance frameworks.”

Real-time visual intelligence
Computer vision—AI that interprets visual information continuously—enables live monitoring and analysis. Companies like LiveSwitch now offer AI that “analyzes videos and photos to automatically generate inventory lists, job measurements” for home service professionals.

Community AI networks
Shared knowledge systems connecting neighbors and local experts, though these remain mostly conceptual rather than deployed (the infrastructure challenges are significant).

Implementation frameworks
LangChain and AutoGPT—frameworks for building AI agents—enable persistent home intelligence, though “workflows requiring persistent memory or autonomous decision-making highlight the platform’s limitations” due to stateless architecture.

This represents the current reality mixed with future possibilities. Some people are building these systems now, though they require significant technical setup and maintenance (and patience when they break).

What I’m actually learning

My moving experience confirmed that systematic visual AI workflows deliver practical results. It also revealed I’m probably working at the easier end of what’s possible (which is fine by me for this topic).

The future likely involves AI systems that understand your space thoroughly enough to predict needs before you recognize them. Some of our readers might already be building toward this future.

For most of us, systematic visual workflows provide a solid foundation. Start simple. Build consistently. Learn from what works.

Vector databases—specialized storage systems that help AI quickly find relevant information—power much of this functionality. When you upload photos and create queries, AI uses these databases to match your situation with relevant solutions.

Where do you stand?

Keep building,

Mark
The AI Learning Guy
👋⚡😎