A study attempts to measure the "alternative boundary": the length of tasks that agents succeed doubles every 129 days, but market signals remain mixed.
With programming agents, could a company build the tools it has been buying until now? Ripon Chandra Malo, Zimeng Yu and Tong Qiu define a "substitution frontier": the set of functions that a customer can develop and maintain at an acceptable level for less than a subscription. They compare it with public data on the capabilities of the agents and on the software market.
Accelerating capacities, a resilient market
In terms of capabilities, the length of tasks (measured in human work time) that the best agents succeed on every other occasion has doubled about every 129 days since 2023. The authors also identified 575 potential open source alternatives to commercial products, and observed a suite of tools, Craft, produce 12 applications in twelve days of public development — early versions whose reliability remains to be established.
On the market side, the picture is mixed: the share of purchased software in U.S. investment has increased and the number of commercial installations tracked has increased. But 32% of survey respondents say they have already given up on at least one purchase because their organization could develop the feature with agents. The authors see this as a warning to publishers that individual purchasing decisions could tip over long before a fundamental movement becomes visible in the numbers.
Read the full article on the original source
Go to the source →