1 min read
Add as a preferred source on Google

Cognition Co-Founder Says Many Software Engineering Tasks Are 'Intelligence Saturated'

Russell Kaplan said cost and speed matter most once a model can perform a task well. He said an analysis he was responding to was consistent with Cognition’s internal data.

An open laptop and mechanical stopwatch sit side by side / TokenPost.ai
An open laptop and mechanical stopwatch sit side by side / TokenPost.ai

Russell Kaplan, Cognition’s president and co-founder, said Oct. 19 that many software engineering tasks are “intelligence saturated,” making cost and speed decisive once a model can perform the work well.

Cognition makes Devin, an AI agent designed to handle software engineering tasks.

Kaplan’s argument begins with a condition: a model must already be able to do a particular task well. “Once any model can do the task well, cost and speed are all that matter,” he wrote in his post. His comments concern those tasks, rather than software engineering work as a whole.

Kaplan also wrote that an analysis he was responding to was “directionally consistent” with Cognition’s internal data. That remark referred to the analysis, while his cost-and-speed argument described how he thinks models should be compared after they can handle a task.

For a task that meets that threshold, Kaplan’s view places less weight on additional model capability and more on the time and cost of getting the work done.

Simon Yoon

Reporter

Simon Yoon reports on blockchain technology for TokenPost. Send corrections or tips to info@tokenpost.com.

Loading…