paperarXivTrust 82 · PrimaryPublished 1mo agoLive · 1mo ago
Cheap Code, Costly Judgment: A Case Study on Governable Agentic Software Engineering
Generative AI is shifting software engineering from a practice organized around scarce implementation effort toward one organized around abundant, low-cost code production. This shift changes the central engineering problem: not whether AI can generate useful code, but how engineers organize architectures, tools, evidence, and feedback loops so that AI-mediated development remains inspectable, correctable, and maintainable. We study this problem through a first-person case study: a 12-week development effort in which a single expert software engineer used frontier AI coding agents to build a
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- LinkedLinked via unknownOpen-source agent framework crosses 50k stars →
- LinkedLinked via unknownPatronus AI lands $50M to build ‘digital worlds’ that stress-test AI agents →
- LinkedLinked via unknownAI coding agents taught robots how to install GPUs and cut zip ties →
- LinkedLinked via unknownClaude Code costs up to $200 a month. Goose does the same thing for free. →
- LinkedLinked via unknownReflections on Software Engineering in the Age of AI →
- LinkedLinked via unknownVibe Coding / Agentic workflow →
- PossiblePossibly related (embedding) · 55%jonwiggins/optio →
- PossiblePossibly related (embedding) · 53%closedloop-ai/claude-plugins →
Covers
newsOpen-source agent framework crosses 50k starsnewsPatronus AI lands $50M to build ‘digital worlds’ that stress-test AI agentsnewsAI coding agents taught robots how to install GPUs and cut zip tiesnewsClaude Code costs up to $200 a month. Goose does the same thing for free.newsReflections on Software Engineering in the Age of AI
Covers (incoming)
newsVibe Coding / Agentic workflownewsDecypher: A Deep Semantic Code Graph for Agentic Coding and Engineering (Now in Beta)newsThe 26.1 Percent Problem: Why One Independent Researcher Started Grading Every AI Skill on GitHub - Programming InsidernewsA randomized trial by METR found that experienced developers completed real coding tasks 19% slower when allowed to use AI tools — yet afterwards, they estimated on average that AI had made them 20% faster. - ScienceBlog.comnewsPluggable by design: An agent mesh for software modernization that adopts the next model releasenewsThe real bottleneck for AI agents may be proving who they arenewsThe cleanup trap: Stop asking RAG to fix bad datanewsHow AI drove Shopify back to clean codenewsWhy good AI agents still produce bad system outputs
Implements (incoming)
repojonwiggins/optiorepoclosedloop-ai/claude-pluginsrepopotpie-ai/potpierepospring-projects/spring-airepoProductive-Superintelligence/lllmrepoNeuralInverse/neuralinverserepoSakshxm1/hermes-agency-orchestratorrepoGreen-PT/honey-for-devsreposammcj/mcp-devtoolsrepoalibaba/spring-ai-alibabarepoarthurpanhku/dvalincoderepodyoshikawa/rulesyncrepoggwhite/4xrepomajiayu000/harnessrepo7-e1even/learn-agentrepoSATYAM-PRATIBHAN/arcmindAIrepoAgent-Field/pr-afrepoTechMitten/Tips-Tricks-for-Agentic-Coderrepologanthorneloe/ai-for-software-engineersrepospec2ship/spec2shiprepozhu1090093659/spec_driven_developrepotruffle-ai/dextorepowindagency/valora.airepoWecoAI/aidemlrepopatchy631/ai-engineering-hubrepostas00/ml-engineeringrepocobusgreyling/loop-engineeringrepoFosowl/agenticSeekrepoNirDiamant/Prompt_Engineering
authored (incoming)
Related across the graph
newsHow AI drove Shopify back to clean coderepoSakshxm1/hermes-agency-orchestratornewsDecypher: A Deep Semantic Code Graph for Agentic Coding and Engineering (Now in Beta)repoProductive-Superintelligence/lllmrepoggwhite/4xpersonKirsten A. DavisnewsThe real bottleneck for AI agents may be proving who they arerepo7-e1even/learn-agentrepoAgent-Field/pr-afnewsThe 26.1 Percent Problem: Why One Independent Researcher Started Grading Every AI Skill on GitHub - Programming Insiderrepotruffle-ai/dextorepojonwiggins/optioreposammcj/mcp-devtoolsrepostas00/ml-engineeringrepozhu1090093659/spec_driven_developrepoNeuralInverse/neuralinverserepopotpie-ai/potpierepomajiayu000/harnessnewsAI coding agents taught robots how to install GPUs and cut zip tiesnewsOpen-source agent framework crosses 50k starsrepopatchy631/ai-engineering-hubpersonTanmay SinglarepoSATYAM-PRATIBHAN/arcmindAIrepoTechMitten/Tips-Tricks-for-Agentic-CoderpersonBerk ÇakarrepoFosowl/agenticSeekrepospec2ship/spec2shipnewsPatronus AI lands $50M to build ‘digital worlds’ that stress-test AI agentsrepospring-projects/spring-ainewsVibe Coding / Agentic workflownewsPluggable by design: An agent mesh for software modernization that adopts the next model releasenewsA randomized trial by METR found that experienced developers completed real coding tasks 19% slower when allowed to use AI tools — yet afterwards, they estimated on average that AI had made them 20% faster. - ScienceBlog.comrepowindagency/valora.airepoGreen-PT/honey-for-devsnewsClaude Code costs up to $200 a month. Goose does the same thing for free.repodyoshikawa/rulesyncrepoWecoAI/aidemlpersonJames C. Davisrepologanthorneloe/ai-for-software-engineersrepoarthurpanhku/dvalincoderepoNirDiamant/Prompt_EngineeringpersonPaschal C. Amusuorepoalibaba/spring-ai-alibabarepoclosedloop-ai/claude-pluginsrepocobusgreyling/loop-engineeringnewsReflections on Software Engineering in the Age of AInewsWhy good AI agents still produce bad system outputsnewsThe cleanup trap: Stop asking RAG to fix bad data