I’ve hand-coded InvMon up to version 1.1. That’s 4 years of conventional, old school, pre-AI coding. Later in 2025 I started to experiment with Claude Code – some of that made it into version 1.2.1. Since November 2025, I’ve been using Claude Code extensively for improvements and new features.
AI coding agents have made incredible progress. Claude Code (in combination with the then current Opus model) made a huge jump in late 2025. The shift of quality that happened around that time had a “life-changing” impact on the way I develop software and work with code.
InvMon is moving forward hand in hand with with agentic coding. While on some projects outside of InvMon I delegate 95% of the work to agents, for InvMon, that percentage is less. I know InvMon’s code-base down to the last line and review every AI change in detail, making sure it fits well into the existing code.
It’s exciting to work with agentic coding. That’s not only because it takes less time now to implement new features but also because it helps to improve the quality of the code-base. Big static code analysis tasks can be performed almost for free. Latent bugs can be identified. Tests can be improved and extended. I can spend more time on overall feature design and improving the user experience and leave the implementation to the LLM.