Linha de pesquisa: Engenharia de Software

Resumo: AI agents’ strategic interactions in economic settings—a key line of inquiry in AI-agent economics—have attracted increasing attention amid recent advances in generative AI. We investigate blockchain-based SWE-Agent outsourcing markets, which link intelligent software engineering and software-engineering economics through programmable, transparent, and auditable mechanisms that centralized platforms rarely expose. We evaluate whether a blockchain-native SDK for economic network simulations can support controlled, paired baseline–intervention experiments under fixed conditions in decentralized SWE-Agent outsourcing markets. We present SWEChain-SDK, a configurable, extensible blockchain-native SDK that provides a local sandbox chain for SWE-Agent software outsourcing auctions and eventcomplete logging of bids, allocations, payments, and artifacts, together with configuration metadata sufficient to support re-runnable experiments under comparable conditions. Using SWEChain-SDK, we conduct paired simulation experiments that vary one policy dimension at a time—(A) comparative advantage via agent specialization, (B) competitive bidding via bidders’ price-signal utilization, and (C) principal-weighted selection via quality-signal utilization—while holding rules, datasets, random seeds, time base, and the agent pool fixed. We report the observed paired differences in task completion, outsourcing cost, and revenue concentration, making trade-offs explicit. The results show that SWEChainSDK supports controlled baseline–intervention experiments on decentralized SWE-Agent outsourcing markets. We contribute (i) a conceptual and terminological framing of SWE-Agent Economics as a domain linking intelligent software engineering and software-engineering economics; (ii) SWEChain-SDK, a blockchain-native SDK for controlled studies of decentralized, auctionbased SWE-Agent outsourcing markets under comparable conditions; and (iii) a public, evaluation-ready release of the SDK—including simulation code, configuration manifests, and event logs—for further empirical studies of decentralized SWE-Agent outsourcing markets.

Link para a defesa: https://teams.microsoft.com/l/meetup-join/19%3ameeting_NWYzY2UyNjUtNjhiY...