Anthropic and OpenAI: Resilience in the Face of AI Instability

The Instability of AI Giants: A Challenge for Businesses

Major research labs and tech giants, often referred to as hyperscalers, promise innovations that could transform our world. While these promises are often kept, they are marred by stability issues, as evidenced by recent policy changes at Anthropic and security incidents at Hugging Face and OpenAI. This instability is concerning for businesses that rely on these technologies for their daily operations.

Companies that base their activities on paid AI models like Anthropic or OpenAI must be able to rely on the reliability of these solutions. However, the current situation shows that this stability is not guaranteed, posing a significant risk for these organizations.

The Rise of Open-Source Solutions

In this context, open-source alternatives like OpenClaw and DeepSeek are beginning to stand out. These solutions offer AI models of comparable quality but without the high costs associated with paid models. Furthermore, the barriers related to usability, security, and accessibility, which could have hindered businesses, are rapidly diminishing.

This evolution prompts technical leaders, such as CTOs and CIOs, to ask crucial questions: how to maintain reliability while reducing costs, especially when the prices of hyperscaler models are unpredictable?

Towards an Agile and Adaptable Infrastructure

The answer to these questions is not simple, but it is clear that businesses must develop systems that allow for a rapid transition between different models and an adaptation of AI agents' interactions with users and data. By building a solid foundation, organizations can remain agile and keep pace with technological advancements without succumbing to the pressures of the latest trends.

The Hidden Costs of "Tokenmaxxing"

Technical leaders have learned that aiming for long-term stability is wiser than chasing costly trends like "tokenmaxxing," which can drain financial and human resources. Some companies, like Meta, are redirecting their efforts towards strengthening team culture in engineering, seeking to improve morale while reducing internal competition related to token usage.

The goal is to modernize the infrastructure to enable sustainable experimentation with large-scale AI tools, rather than rushing into every new technological trend.

A Collaborative Future for AI in Business

The future of AI in business relies on a harmonious collaboration between agents and humans. This involves recognizing that agents possess capabilities similar to those of humans, with the added advantage of being able to test real infrastructures at scale.

It is crucial to ensure that agents have the same protections as human teams, particularly regarding identity and permission management, and to ensure that all actions are visible and auditable.

Adapting Architecture to New Models

Resilience begins with the acceptance of the constant change of models and their usage. Technical leaders must be ready to adopt the latest models from hyperscalers or integrate innovative open-source models that alter the collaboration between agents and humans.

Agents must be able to interact with real infrastructures and actively participate in business processes, which requires a level playing field between them and engineers.

Flexibility as a Major Asset

The era when workflows and knowledge were locked within a single provider is over. Today, flexibility, elasticity, and adaptability are essential for building systems and teams capable of adjusting to new market conditions.

Companies must avoid long-term lock-in and prepare to integrate new models and tools without having to rebuild their infrastructure. This flexible approach allows for building sustainable resilience in the face of technological changes.

Technical leaders must adopt this strategy now to ensure the longevity and efficiency of their operations in an ever-evolving technological landscape.