Should firms build local, script-based tools instead of AI-dependent tools to future-proof against rising AI costs?

The concern raised was that token costs and hardware costs for running AI (including local models) are likely to become significant over time, so it may make sense to build local script-based tools now while AI assistance is relatively cheap. The response affirmed this: you build with AI, but what you're building is not necessarily AI itself. If you produce tightly organized scripts or tools using AI as the builder, those artifacts are yours forever — they'll keep working even if the underlying LLM service changed or became unavailable, as long as the systems around them still function. A distinction was drawn: tools like HTML-based utilities that don't call out to an API generally aren't dependent on ongoing AI costs, whereas tools that require an API connection do incur ongoing AI-related costs and dependency. The overall takeaway was that now is a good time to build and explore, since AI-assisted development is likely to get more expensive later.