Anthropic claims half of its agent tool calls are to do with software engineering - so are developers letting AI take over?
Anthropic's AI agents are increasingly used by developers
Sign up for breaking news, reviews, opinion, top tech deals, and more.
You are now subscribed
Your newsletter sign-up was successful
- Claude is used far more by developers than customer services, sales, finance or ecommerce
- It's now working autonomously more than it ever did, including complex tasks
- Claude clarifies details more than humans interrupt it
Anthropic has revealed around half of all Public API tool calls come from software engineering, with other sectors such as customer service, sales, finance and ecommerce only accounting for a few percentage points each.
Alongside this, Claude Code was found to be working autonomously for longer than ever before, now at over 45 minutes compared with under 25 minutes just three months ago.
OpenAI says the increase in autonomy doesn't just relate to its latest, high-performance models, suggesting a broader shift to agentic AI rather than trials of its newer models.
Agentic AI is here, and it's most common in coding
In a new blog post, OpenAI outlined how users are starting to build trust with AI as tasks become more ambitious, however there's still a 'deployment overhang' whereby models are still more capable than users currently allow. External evaluations also prove that models can handle much longer tasks in ideal conditions.
Claude is also helping on the trust front, pausing to ask clarification questions more than humans interrupt it.
In terms of coding, it seems that the more complex the task, the less human oversight there is. Complex ones like exploit discovery and compiler writing see around 67% human oversight, compared with 87% oversight on simple API tasks.
Claude also asks around twice as many questions on high-complexity tasks than simple ones.
Sign up to the TechRadar Pro newsletter to get all the top news, opinion, features and guidance your business needs to succeed!
Looking ahead, Anthropic covered the importance of training models to recognize uncertainty and to proactively ask clarifying questions. The company also discourages mandating manual approval for every action, claiming that it adds friction without necessary boosting safety.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!
And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.
With several years’ experience freelancing in tech and automotive circles, Craig’s specific interests lie in technology that is designed to better our lives, including AI and ML, productivity aids, and smart fitness. He is also passionate about cars and the decarbonisation of personal transportation. As an avid bargain-hunter, you can be sure that any deal Craig finds is top value!
You must confirm your public display name before commenting
Please logout and then login again, you will then be prompted to enter your display name.
