The AI Pro
A collection of 36 posts
The AI Pro
If you find yourself struggling to get good results with your AI tools, this workshop is for you. Learn critical AI skills so you can stay on top of this industry.
· Rob Conery
FREE: You're Ready. Now Go Build Something.
You've got the techniques. You've got the tools. The only thing left is to use them. Go make something real.
· Rob Conery
LAB: ETL With Claude, Fake Data, and a Sci-Fi Obsession
Generate fake data, validate it, and analyze it, using Claude to write every script. The prompt engineering trick is the point.
· Rob Conery
Transform Messy Data Without Writing Transformers
ETL is tedious. Mapping fields, handling edge cases, validating output. Let the AI do the boring parts while you focus on what matters.
· Rob Conery
Skip the Pandas Tutorial
You don't need to learn a data analysis library to answer questions about your data. Describe what you want and let Claude write the code.
· Rob Conery
FREE: The Laziest Way to Automate Anything
Don't ask AI to do repetitive work directly. Ask it to write a script, then run that script as many times as you want for free.
· Rob Conery
Understanding the Context Window (and why Claude Halucinates)
This is one of the core parts of your job! You have to understand how you're billed, and how Claude can do its work the best.
· Rob Conery
FREE: Understanding Token Use
Tokens are money. A few bad habits can burn through your cap in one session. A few good ones can stretch it for weeks.
· Rob Conery
LAB: Path Instructions, Agents, and Skills
Build layered Claude Code context using path-based CLAUDE.md files, create a custom agent with the generator, and use metaprompting to build a GitHub skill from scratch.
· Rob Conery
Sharing Instructions and Context Across Agents with Skills
Skills let you package complex workflows into a single command. Build once, run anytime, and stop repeating yourself.
· Rob Conery
Isolated Tasks, Smaller Context with Agents
Claude's subagents fire automatically based on your prompt. Less manual than Copilot's agents, which means less control but also less overhead.
· Rob Conery
Claude Code Instructions
Claude can pick up context based on where you're working in your project. A bit noisier than Copilot's approach, but just as powerful.
· Rob Conery
Catch Obvious Problems Before PR Review
AI-generated code can have subtle issues. A quick sanity check catches the embarrassing stuff before your teammates do.
· Rob Conery
Prompts You Can Reuse Forever
Your best prompts shouldn't live in your chat history. Refactor them into reusable assets that compound your investment over time.
· Rob Conery
An Agent That Catches Mistakes Before You Do
An agent doesn't have to be a persona. It can be a process, like test-driven development, that runs automatically and keeps you honest.
· Rob Conery
Build a Specialist That Knows Your Stack
Stop re-explaining your tech stack every session. Custom agents remember your context so you can skip straight to the work.
· Rob Conery
Creating Instructions for Specific Needs
Custom instructions are cool. Path-based instructions are cool. You know what's cooler? Instructions that target a very specific need, like service logic!
· Rob Conery
Different Rules for Different Parts of Your Codebase
Your tests need different guidance than your services. Path-based instructions let you give the AI context that matches where you're working.
· Rob Conery
Keeping the Receipts with Session Recaps
It's not enough to fire up a planning document and let Claude do it's thing. You have to be able to answer for what was done! That's where recaps come in.
· Rob Conery
The Flow Montage: Running Our Iterations
We have liftoff! Let's get this POC rolling by jamming out our iterations from our planning doc. This will be "montage style", where we only stop if something fun happens.
· Rob Conery
POC Iteration 1: Getting Into The Flow
AI sessions drift. Features creep. An iteration plan keeps both you and your AI pointed at the same target, even on multi-day projects.
· Rob Conery
Build Something Real Before You Build Everything
Big projects fail when you commit to the wrong approach too early. Use AI to build a small, working slice first and validate before you go all-in.
· Rob Conery
Thinking About Deployment Upfront
One of the things I've learned over the years is to know where your application is going to live during production. How do you make that choice?
· Rob Conery
FREE: Stop Staring at a Blank Cursor
Not sure what to ask? Ask the AI to help you figure it out. This simple technique eliminates prompt paralysis and gets you moving.
· Rob Conery
Claude's Power Features
Claude Code has capabilities that aren't obvious at first glance. Learn what's there so you can decide what's worth using.
· Rob Conery
The Copilot Features You're Probably Ignoring
Most people use 20% of what Copilot offers. These settings and features will cut the noise and make the tool work the way you think.
· Rob Conery
Let Claude Trace the Problem for You
Claude Code gives you debug logs and telemetry you can actually control. Less visual than Copilot, but more flexible when you need to dig deep.
· Rob Conery
Find Bugs Without printf Debugging
When the AI gives you nonsense, you need to see what's actually happening. Copilot's debug view shows you exactly how your prompt gets processed.
· Rob Conery
FREE: Commits That Explain Themselves
Go beyond basic commit messages. Create a reusable prompt that generates commits your whole team will appreciate.
· Rob Conery
Stop Dreading Git Commit Messages
You know the feeling: staring at the commit message box, trying to summarize what you just did. Let the AI handle this daily friction.
· Rob Conery