From vision to code: Building an AI toolkit

People often ask me, “Hey, Reiner, what exactly are you doing with AI?” It’s a tricky question to answer directly, because I don’t start with solutions—I start with listening. When someone tells me about a tedious task they’re facing or a process they wish worked better, that’s when the ideas start flowing. I get this hunch about how we might solve it, and that’s where the journey begins.

In my previous article “My AI Journey (So Far): Real Projects, Real Impact”, I shared my thoughts on AI as a tool for empowerment rather than replacement. Today, I want to show you what that journey looks like in practice—the actual tools and scripts I’ve been building along the way.

This list isn’t about showing off finished products. It’s proof of my approach: try things out, see what works, improve what’s promising, and don’t be afraid to scratch ideas that don’t quite hit the mark. I’m especially keen to work with people who want to experiment, who are willing to try new approaches even if they’re not perfect yet. Every tool here started with a conversation and a “what if we tried this?” moment.

Moving beyond ChatGPT

Like many of you, I started with the usual ChatGPT prompts and experiments. But as I dug deeper into AI’s possibilities, I found myself facing specific challenges that needed solving. You know how it goes—you’re in the middle of work, and you think “there must be a better way to do this.” That’s exactly how each of these tools came about.

Some might sound a bit obscure at first—like a script that checks links in markdown files or converts European prices to AUD. But each one exists because I hit a real roadblock in my work that needed solving. Whether it was spending hours manually checking if website links still worked, or trying to properly document foreign expenses for the ATO, these weren’t theoretical problems—they were actual time-sinks in my day-to-day work.

So I started building custom applications that could handle these specific challenges in content creation, data analysis, and workflow automation. These aren’t groundbreaking innovations—they’re practical tools built to make day-to-day tasks easier and more efficient. Each one scratches a particular itch, solving a real problem that I (and maybe you too) have faced.

Let me walk you through some of the key projects I’ve been working on, organised by their primary focus areas:

Content creation and analysis

The content space has certainly changed with AI, but I wanted to go beyond simple text generation. I’ve built tools that help optimise and validate content while maintaining that crucial human expertise:

• A prompt improvement system using DSPy that helps refine AI interactions
• A keyword analysis tool focused on the Australian market
• Tools for managing markdown documents, frontmatter, and content integrity
• A sync system between Obsidian notes and Hugo websites

Local data analysis

Understanding that overseas tools often don’t quite fit our needs here in Australia, I’ve developed several local applications:

• An SEO performance analysis tool that combines data from multiple sources
• A finance manager designed for Australian tax requirements and GST
• An ATO-compliant EUR to AUD currency converter using RBA rates

Research and automation

Some of my more interesting projects involve creating systems that can understand and process information at scale:

• A webcrawling research assistant for gathering and analysing information
• An email response system with a custom knowledge base
• Perplexica—a search engine combining multiple AI models
• A RAG hallucination checker to help keep AI responses factual

Under the bonnet

These tools combine various technologies including:

• Local and cloud-based LLMs
• Custom RAG systems
• Automated data processing
• Web scraping and content analysis
• Integration with official Australian data sources

Where to from here?

These tools are all works in progress—some quite basic, others more developed. They represent my ongoing effort to create practical AI implementations that actually help with real-world tasks, particularly focused on Australian business needs.

In line with my previous article’s emphasis on human-centred AI, each tool is designed to handle the repetitive bits, leaving humans free to focus on strategy and creative thinking.

Want to know more?

All these tools are still evolving, and I’m always keen to improve them. If you’re interested in any particular tool or would like to collaborate on making them better, give me a shout. The code is available to examine, and I’m happy to discuss how these tools might work for different use cases.

Whether you’re looking to streamline your content workflow, improve your SEO analysis, or build smarter research tools, there might be something in this toolkit that could help—or at least spark some ideas for your own AI projects.

Keen to learn more about any of these tools or chat about how they might help with your specific needs? Let’s chat