MC.98: The Software-Eating AI: Why Anthropic's Opus is Redefining Work

"If software is eating the world, Anthropic is eating software." This week, I want to talk about a significant development that I believe profoundly changes how we'll interact with technology and how work gets done: Anthropic's Claude Opus.

Having spent some time delving into the latest iteration of Opus, my primary takeaway is clear: Opus is deeply oriented towards code, and in an increasingly digital world, everything is becoming code.

Anthropic isn't just releasing another powerful model; they're strategically pushing the boundaries of code execution, a move that, while subtle, has massive implications. Two recent announcements perfectly illustrate this direction:

1. Advanced Tool Use: This powerful feature allows Claude to dynamically discover, learn, and execute tools. Imagine an AI that can intelligently navigate vast tool libraries, calling upon only the necessary functions for a given task. This drastically reduces the context window burden and significantly enhances accuracy for complex, multi-step workflows, moving beyond simple function calling to sophisticated, programmatic orchestration.

2. Claude For Excel: This integration brings Claude directly into your spreadsheets. It's not merely a chatbot; it's an intelligent agent capable of understanding entire workbooks—from intricate nested formulas to multi-tab dependencies. It can debug errors, test scenarios, and even build models, all through intelligent code execution directly within Excel.

These aren't merely incremental updates; they represent a concerted effort to empower AI to perform "real work" by leveraging code at its core.

We are witnessing a fundamental paradigm shift. Instead of humans learning the intricacies of various software applications, we're moving towards a model where a request is made, and an AI constructs the entire artifact by abstracting and executing code. This means less time spent on manual operations and more focus on strategic outcomes.

The Strategic Power of Opus Pricing

The most telling sign of Anthropic's strategic intent, beyond the technical capabilities, lies in their innovative pricing for Opus. Historically, high-tier models like previous Opus versions (e.g., Opus 4.1's Input: $15/MTok, Output: $75/MTok) were often cost-prohibitive for widespread adoption in daily workflows.

However, the new pricing for Opus 4.5 dramatically changes the landscape, bringing it down to $5/million tokens for input and $25/million tokens for output. This isn't just a discount; it's a strategic realignment that positions Opus as a highly competitive and “accessible” model.

This aggressive pricing strategy, bringing Opus's cost significantly closer to models previously considered more affordable, signals a clear message from Anthropic: they intend for Opus to become the go-to model for a vast range of tasks. It makes the cutting-edge capabilities of Opus—especially its proficiency in code and agentic workflows—available to a much broader audience, from individual developers to large enterprises.

This pricing shift, coupled with features like Advanced Tool Use and Claude for Excel, indicates that Anthropic sees code execution as the future of AI-driven productivity. They are making a bold move to commoditize high-performance, code-centric AI, establishing it as the new baseline for how we interact with software and produce tangible results.

We're moving beyond a simple "bigger is better" model race. The future is about smarter integration, efficient execution, and specialized agents that transform raw instructions into finished products through the elegant power of code.

Until next Thursday 🎉
Olivier

Bonus: I love when the model thinks outside the box. Instead of using code execution and Python libraries to create a PDF presentation, Opus simply uses a headless browser to save an HTML presentation as a PDF. Genius

Like this newsletter? Forward it to a friend and have them sign up here.

Until next Thursday 🎉