When I first started using Agentic LLM coding assistants, I was thrilled by their potential, but also often frustrated by their unpredictability.
Some days, they produced near-perfect code; other times, inexplicable errors crept in, despite similar instructions. In fairness, sometimes it was me and sometimes it was them.
It didn’t take long to realize that AI coding isn’t just about the model ... it’s about how you interact with it ... what tools you use ... and it's also sometimes influenced by factories completely outside of your control.
📝 Precision in Prompts and Rules: The quality of your prompts and the clarity of system-level instructions drastically impact output. Structured approaches like SPARC excel because of their robust rule sets, precise directives and iterative build/test/reflect/improve loops. Some agents (like Replit) try to incorporate these best practices behind the scenes, improving the quality of the output. Tools like Cursor and Augment Code work hard to refine system-level prompts and augment baseline models, ensuring more consistent and effective results.
🤝 Managing Development Dialogue: Engaging thoughtfully with the agent throughout the entire development process is essential. Providing clear instructions, encouraging iteration and refinement, asking it to evaluate its own work, and structuring dialogue thoughtfully, help maintain quality, even when external factors cause performance fluctuations. Don't rush. Be precise. Make it test it's own work.
⏳ Timing and System Load: Quality can vary based on time of day and server demand. In Eastern Time, mid-late afternoons and even occasionally weekends (vibe code mania) seem to bring performance dips, likely due to higher loads and resource allocation ... while late evening sees a rebound. Recognizing these patterns allows for smarter task scheduling. If switching tools doesn't help, take a break, go for a walk, listen to a podcast, do some product research on Perplexity ... even do something analog 😀
🔄 Keeping Up with Platforms: Staying updated on the latest LLM versions is crucial. The capability difference between models like Claude 4 vs. Claude 3.5, or o3 vs Gemini 2.5 Pro WRT diverse problem sets, is material. Some models are clearly better at some things than others. And knowing when to request deep thinking is important (certainly from a cost benefit basis). Get the model to document their work so they have a long-term record of what they have done and what they have been thinking. This really helps with limited context windows.
So, while LLM variability presents challenges, optimizing prompts, structuring interactions, being strategic in timing, and leveraging the right tools can significantly improve results.
Welcome to the entropy of emergent systems.
Mark Ruddock
Thoughts of an internationally experienced growth stage CEO and Board Member.
Thoughts of an internationally experienced growth stage CEO and Board Member
Mark Ruddock
Thoughts of an internationally experienced growth stage CEO and Board Member.
Experienced international CEO with three exits and over 20 years of experience at the helm of VC-backed technology and fintech startups. I've founded companies and come in as a later-stage CEO to help existing companies scale. My experience spans multi-country online financial services, mission-critical enterprise software and consumer-focused mobile applications.
Currently CEO of Kontent.ai, the industry's first CMS with native AI capabilities, kontent.ai helps the world's leading organizations generate an unparalleled return on their content by enabling them to create, manage, distribute, and optimize their content at scale with unprecedented speed oversight, and security.
Before that, I was CEO of Nuula (formerly known as BFS Capital), which Nav acquired. At Nuula, we believed there was a better way to help run your small business — so we built it. Nuula is a mobile application that gives small business owners instant access to critical business metrics and innovative financial products anytime, anywhere. It provides small business owners with the tools, content, and capital they need to be successful in a modern, competitive market.
Prior to that interim CEO of 4finance. Agreed to take on an interim role to help clarify strategy, accelerate the critical transformation of the business and set the company on a course for the next growth phase.
Before that, I was Managing Director, International at Wonga, responsible for overseeing all businesses outside of the UK, including consumer lending businesses in Canada, Spain, Poland and South Africa and our eCommerce and product financing business BillPay in Germany. Prior to that was CEO of Wonga Canada.
Prior to that, I was EO Viigo. When Viigo was acquired by Blackberry in 2010, the company had hundreds of partners, won numerous awards, including “PC Magazine’s Editors Choice“, the “Most Promising New Company of the Year” from the CNMA, and the “Wireless Leadership Award” from RIM, signed marquee enterprise customers such as Oppenheimer, UBS, BAT and Scotts and been downloaded over 3MM times, becoming one of the most popular BlackBerry apps of all time.
I started my career as the Founder and CEO of INEA. A software company that was a strategic partner to many of the world's leading financial institutions, INEA had deployments in over 44 countries. Backed by RBC, Ventures West, APAX and Edgestone, the company was acquired in June 2005 by Cartesis. Cartesis was subsequently acquired by Business Objects in 2007, which was later acquired by SAP.
AFINEA | 136 Arundel Ave, Toronto