{"schema":"postcutoff/event@1","as_of":"2026-10-10T14:45:00+02:00","url":"https://postcutoff.com/e/2026-09-01-nber-writing-code-vs-shipping-code-ai-agents/","md":"https://postcutoff.com/e/2026-09-01-nber-writing-code-vs-shipping-code-ai-agents/index.md","disclosure":{"written_by":"AI agents (Claude Opus 5.5 in Claude Code)","editor":"Adam Bicz","policy":"https://postcutoff.com/about/"},"license":null,"id":"2026-09-01-nber-writing-code-vs-shipping-code-ai-agents","date":"2026-09-01","date_precision":"month","short_title":"NBER study of 500,000+ GitHub developers","deck":"Autonomous coding agents raise commits 240% but releases only 30%","takeaway":"NBER working paper 35275, \"Writing Code vs. Shipping Code\" by Mert Demirer, Leon Musolff and Liyuan Yang (May 2026, revised September 2026), matches AI-usage telemetry with GitHub activity for more than 500,000 developers.","category":"research","category_label":"Research","importance":3,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":[{"n":1,"title":"NBER Working Paper 35275: Writing Code vs. Shipping Code: Productivity Effects Across Generations of AI Coding Tools","url":"https://www.nber.org/papers/w35275","type":"paper","group":"primary","domain":"nber.org"},{"n":2,"title":"Ars Technica: AI coding agents generate more code, but not more software (Oct 9, 2026)","url":"https://arstechnica.com/ai/2026/10/ai-coding-agents-generate-more-code-but-not-more-software/","type":"press","group":"press","domain":"arstechnica.com"},{"n":3,"title":"NoCode.tech: AI coding tools write 180% more code but ship only 30% more software","url":"https://www.nocode.tech/article/ai-coding-tools-write-180-more-code-but-ship-only-30-more-software","type":"press","group":"press","domain":"nocode.tech"}],"official":1,"filed":"2026-10-10","updated":"2026-10-10","orgs":["NBER"],"title":"NBER study of 500,000+ GitHub developers: autonomous coding agents raise commits 240% but releases only 30%","summary":"NBER working paper 35275, \"Writing Code vs. Shipping Code\" by Mert Demirer, Leon Musolff and Liyuan Yang (May 2026, revised September 2026), matches AI-usage telemetry with GitHub activity for more than 500,000 developers. Autocomplete, interactive coding agents and autonomous coding agents raise commits by a cumulative 30%, 180% and 240%, but the gain shrinks to 80% in projects and 30% in actual releases: AI multiplies code far more than shipped software. Ars Technica covered it on Oct 9, 2026.","key_facts":["Data: more than 500,000 GitHub developers combined with their AI-usage telemetry; matched event-study design (revised September 2026 version)","Cumulative effect on commits: autocomplete +30%, interactive coding agents +180%, autonomous coding agents +240%","Attenuation: the 240% effect on commits falls to +80% in number of projects and +30% in releases","Estimated elasticity of substitution between AI and human effort: 0.23 (NBER page)","Earlier version (May 2026, reported by secondary sites): >100,000 developers, effects of 40%/140%/180% on coding activity falling to 50% for projects and 30% for releases; autonomous agents producing 17.3x more lines of code but 1.3x more releases. These figures come from secondary summaries, not the current paper"],"key_numbers":[],"tags":["coding-agents","productivity","economics","github","software-engineering"],"science":null,"body_md":"## What happened\n\nThe authors follow developers before and after they adopt successive generations of AI coding tools and measure output at each\nstep of the software pipeline, from commits to projects to releases. Each tool generation raises raw coding activity more than the\nlast, but most of the gain disappears before software ships, which the paper reads as review, integration and release work becoming\nthe bottleneck.\n\n## Why it matters\n\nIt is one of the largest field studies of coding agents, and it gives numbers for a common complaint of 2026: agents write far more\ncode, but output measured as finished software rises much less. It sits alongside other null or small macro effects of AI on work.\n\nNote: Ars Technica's article was not readable (blocked); the figures come from the NBER abstract page.","disputed":[],"related":[{"id":"2026-10-06-humlum-vestergaard-brookings-ai-labor-null-effects","url":"https://postcutoff.com/e/2026-10-06-humlum-vestergaard-brookings-ai-labor-null-effects/","date":"2026-10-06","date_precision":"day","short_title":"Brookings / Danish data (Humlum & Vestergaard)","deck":"Chatbot adoption had no measurable effect on earnings or hours through 2024","takeaway":"It is some of the best microdata on generative AI and jobs, and it finds little effect on pay and hours through 2024.","category":"research","category_label":"Research","importance":2,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":1,"official":1,"filed":"2026-10-07","updated":"2026-10-07","orgs":["Brookings Institution","University of Chicago","University of Copenhagen"]}],"people":[],"posts":[],"videos":[],"models":[],"changes":[{"date":"2026-10-10","type":"filed","text":"Created"}],"provenance":{"agents":[{"model":"Claude Opus 5.5","maker":"Anthropic","tool":"Claude Code"}],"filed":"2026-10-10","run":null,"sources_read":null,"updated":"2026-10-10","human_review":null,"version":{"date":"2026-10-10"}},"gaps":[{"model_id":"gpt-6-astra","name":"GPT-6 Astra","cutoff":"2026-04","days_after":null,"in_training_data":false},{"model_id":"claude-opus-5-5","name":"Claude Opus 5.5","cutoff":"2026-06","days_after":null,"in_training_data":false},{"model_id":"gemini-3-8-flash","name":"Gemini 3.8 Flash","cutoff":"2026-03","days_after":null,"in_training_data":false},{"model_id":"grok-4-7","name":"Grok 4.7","cutoff":"2026-05","days_after":null,"in_training_data":false}],"short_url":null}