Anthropic’s Claude has been the underdog of the AI race.
While OpenAI’s ChatGPT is looked at as a pioneer and DeepSeek and Grok are the disruptors, Claude existed in the space between.
But with its 3.5 Sonnet release, Claude broke many benchmarks. It could reason better and also perform creative tasks better. People had to take note of Claude and, by extension, Anthropic.
However, with DeepSeek, investors and the silicon valley began asking questions.
What are the labs in the US doing? Why is a Chinese model cheap and outperforming such well-funded models?
Anthropic’s CEO, Dario Amodei, believed that the claims were overstated -many questioned him.
But Anthropic returned with an answer: Claude 3.7 Sonnet.
As of today, Claude has outperformed all AI models in programming tests. Even the SWE-bench tests a language model’s ability to solve real-life GitHub problems.
They also announced a CLI tool called the Claude Code. This agentic tool allows developers to “delegate substantial engineering tasks to Claude directly from the terminal.“
But it’s not all work and no play. To show its reasoning capability, Anthropic made Claude play Pokémon. And while the 3.5 did not go as far as they expected it to, 3.7 defeated Brock and Misty— two gym masters.
That means the tool can reason independently and come up with strategic responses to scenarios.
AI Development: Where does it go from here?
The development of AI seemed to affect jobs that were not complex enough. However, the recent changes and the leaps and bounds with which the models have developed have become scary.
And that’s to say the least.
Mathematics, programming, and other complex tasks that involve reasoning can be done by the models. They are improving at a faster rate.
But, if AI is developing its reasoning, that means the learning opportunities open up, and so do the possibilities. Based on this perspective, we might be in the golden age of learning. Especially from a tool that can help us be better at what we do.