Another busy day.
Tested cursor + sonnet 3.7, and the results were pretty good. But there’s still too much content, so we’ll need to fine-tune the rules accordingly. The usual issue with generative AI is that it tends to overthink, often drifting away from the actual problem that needs to be solved.
If we think about it, AI and humans have a lot in common. Since it's still like a child, maybe we should be more patient and refine our prompts accordingly.
Lately, SNS is flooded with posts like "This will be huge", "It's over", and so on—it’s getting a bit tiresome. No need to panic. Performance always depends on specific content (where answers already exist), but real-world problems are usually unique and don’t come with a predefined database.
But occasionally, I see some developers feeling like they’re living in a parallel universe, not experiencing the productivity boost they expected.
Most people on SNS are just trying to grab attention, so it’s normal for content to be exaggerated or detached from real-world applications. No need to panic—just focus on what new models, technologies, and tools can actually help us solve.
Personally, I find cursor quite helpful for understanding projects that have passed through multiple hands. However, in other cases, a 30–40% productivity increase is already pretty decent. We have to factor in the time spent crafting prompts and the extra debugging required. Even if a feature is completed, it’s not something we can apply directly—it still needs a proper code review. And if we include maintenance in the equation, the overall impact on productivity becomes even more limited.
Top comments (0)