Managing Unreliable Compilers
There’s a lot of discussion on X and other places about the future of software development. As with many things in life, the reality is both complex and in the middle of the extreme viewpoints. What we’re seeing at Tonkotsu:
Agents are fast and powerful, but make mistakes. They can’t operate unsupervised. We think they’re like unreliable compilers.
That means developers are as critical as ever, but their role shifts to being managers of coding agents.
This transformation means developers need to be focused on planning and verification, while delegating coding. The role has become barbell-shaped, and the industry needs new tools and workflows to accommodate this.
More here → https://blog.tonkotsu.ai/p/managing-unreliable-compilers
What are you seeing play out in your experience? How is development changing and what do developers need to do to stay on top?
Replies
For myself, I have been using both Windsurf on VSCOde and a local install of deepseek. I have saved hours upon hours just in the autocomplete, in particular for JSONs and translation files. This gives me time to focus more time on the other parts of my development tasks. I find Windsurf particularly useful and I can see expanding into automating documentation. However, there are so so many choices that it took me a long time to find tools that were free and actually worked.
we should emphasize the current limitations of AI models, particularly in context size, economic efficiency, creativity, and hallucination reduction. Until ai can manage a context size of 10-20 million tokens, developers should expect several years before significant changes in their role. Companies relying solely on ai for coding without developers risk facing challenges, as aicannot fully replace human expertise.
ai can assist with tasks like writing code and translating text, it is not a substitute for human oversight. so i expect that a hybrid approach will become the standard, where ai tools augment rather than replace developers. Businesses should be cautious about overreliance on ai without maintaining skilled development teams to avoid potential issues in the near future.
i think the key is a balanced approach, ais as assistants while ensuring that human expertise remains integral to development processes. As models evolve, this balance may shift, but for now, a hybrid strategy i think is the way to avoid the pitfalls of relying solely on ai
Tonkotsu
@cassi_cassi Completely agree that a hybrid/balanced approach is needed. We believe the critical questions to answer with such an approach is: when and how does the human supervise the process? Our take is that developer attention and judgment should concentrate on the endpoints of the process: planning and verification. What's in the middle can be delegated away, but it's important for developers to exercise their judgment on making key technical decisions and holding a high quality bar. So the role is now barbell-shaped.
vibecoder.date
I want to become a former developer. I used to love coding, I still do, through the ups and downs. People sometimes mistake my expertise with AI for passion for how the technology is being packaged. This is not the case.
I think the concept of a developer is going to morph into domain expert with problem solving skills.
Tonkotsu
@build_with_aj Agreed: domain expert + technical oversight seems like the right combination. There's so much to build to re-tune dev tools to that role. I also think it increases the momentum behind the rise of the "product engineer".