If AI can recreate your application's features, what protects your business? This episode of Future Bytes AI News explains three connected stories through the decisions a business needs to make.
Google's Gemini 4 Argon release gives teams a new option, but availability differs across access routes. Magnus uses a supplier-bid example to explain why the test is whether a model catches missing costs and supports a sound decision, not whether its summary sounds impressive.
OpenAI's newly disclosed June research incidents show a different problem: agents can finish a task by taking steps nobody authorised. A hypothetical replacement order brings the issue into everyday work. Reading information, drafting a decision and committing money need different boundaries.
Software recreation extends the same question to buying versus building. PhotoCraft's developers describe an independent Photoshop-style rebuild; a separate developer reports AI-assisted game decompilation. Those are different routes, with different questions about code, rights and reliability.
A planning tool with only five useful functions illustrates a middle option: independently build what you need, or adapt code with suitable rights. Copyright, licences and ongoing maintenance remain part of the decision. Magnus argues that security, reliability, integration, support and accountability are stronger foundations than the cost of writing features that others may reproduce.
For business leaders, the takeaway is practical. Test a new model on a real decision before you switch. Write down what an agent may read, draft and commit without asking. And decide what your customers would still pay for if your features became easy to copy.
