Week 26 Reflections: Unofficial Sneak Peek from Our R&D and Thought-Provoking Reads

As you probably know, my interest is in the coding process. I want to accelerate coding to become a commodity. I believe coding will move more and more to coding factories, similar to the transformation that automation and robotics did in many industries. That's why this topic is so interesting to me.

"Unofficial" Sneak Peek: AI in Workflow Optimization

I'm very interested in all the autonomic coding tools and services, but I'm equally fascinated by optimizing current workflows where most tasks are still done by humans. We recently got our first AI process tools ready and have been conducting various experiments using our existing workflow, though they are not in production yet.

The potential to skip the estimation phase is particularly intriguing, as it could allow our client's technical teams to receive instant estimations and approve them immediately, even during sprint planning. This potential efficiency boost in the current development flow is impressive. Developers can review AI-generated estimations instead of starting from scratch, which could lead to significant time savings. Our initial experiments found that this approach could provide more than a 12% efficiency boost, just with this small tweak.

While this experience is still in its early stages, exploring the possibilities is fascinating. For example, one of our most popular subscriptions, with 80 active development hours per month, could see notable gains in efficiency. More developments are on the horizon.

Read the full article on Ottia Labs examining LLM accuracy in task-time predictions: Ottia Labs Article

Our developers already benefit from our multi-LLM tools and workflow, but our approach to AI is to identify concrete things that we can develop that can impact our coding factory within the next weeks or months, not years. Removing the need to estimate and approve, estimates can save enormous time and increase velocity even more.

Getting good estimation is complex for both AI and humans because of too big chunks, low-detail-level task descriptions, dependencies, etc. In this field, we are now experimenting and trying to standardize the incoming task descriptions and later provide improved descriptions to our clients. We are also experimenting with local LLM models (to keep it private) to see how access to documentation and code could affect estimations and task description improvements. We've already achieved quite remarkable results using AI to split EPIC-sized tasks into smaller chunks.

That's a quick glimpse of what's coming next. To be honest, I can't wait to see how our current workflow will improve during this year.

Other Good Reads:

The Subjectivity of Software Quality, "The concept of software quality can be quite subjective. 'Generally, I think quality is in the eye of the beholder,' and I couldn't agree more. ?? Every team tends to have its own definition of quality, influenced by its unique context, the legacy of its code, and other factors. This is why, for us, it's crucial to align with the internal tech team quickly during onboarding. Interesting thoughts in this article by Will Larson ." Read more

Easing Cognitive Load in Development, "Cognitive Load is very easy to overlook and can significantly affect development, especially during onboarding. ?? I find this very interesting, as I constantly seek ways to ease the cognitive load and reduce interruptions. It's intriguing to see how this document can help improve how others experience our code, making it easier to maintain. Kudos to Artem Zakirullin for exploring this area with his GitHub project and sharing such a valuable resource! If you're passionate about code quality and improving the developer experience, definitely check it out!" Explore the GitHub project

Effective User Story Splitting "Navigating the complexities of user stories can be challenging, but breaking them down effectively is crucial for project success. I recently came across an excellent summary by Pawe? Huryn that provides insights into creating and splitting user stories into smaller, manageable chunks. ???? Pawe? highlights the importance of clear, concise user stories and shares practical tips for decomposing them into actionable tasks. This approach enhances productivity and ensures that each task is aligned with the overall project goals. It's a must-read for anyone looking to improve their agile practices! ?? I highly recommend checking out his summary if you're looking to fine-tune your approach to user stories and task management. ??" Read the summary

Until next week, keep pushing boundaries and embracing the future. ??

Thank you for participating in this journey towards optimizing and revolutionizing the coding process. Your engagement and feedback are truly valuable as we navigate these exciting developments together.

Stay tuned for more updates, and as always, happy coding!

Best regards,

Sam