Why aren't we shipping faster now?
Not investment advice. Muhabbit scores how much a story may matter, not which way to trade.
A new body of 2026 research, spanning a 30-interview study, an NBER working paper (No. 35275) and several arXiv preprints from researchers including Cha, Badea, Bird, Butler and Murphy-Hill, finds that AI coding tools raise measured activity—code written and tasks completed—without a matching gain in shipped software, a gap the authors call the activity-outcome gap. The NBER paper by Demirer, Musolff and Yang compares productivity effects across generations of AI coding tools, while a separate study asks what AI has actually let developers unlock. Other papers examine when code-review agents can run "lights-off," finding limits in bug detection, and where developers draw the line on AI autonomy. The findings matter because firms are spending heavily on AI developer tooling on the premise of faster delivery, yet the evidence suggests output metrics overstate real progress. Next, watch for the 30-interview paper to be released publicly and for follow-up work linking tool use to downstream delivery measures such as release frequency and defect rates.
Muhabbit's AI wrote this summary in its own words from Hacker News's report. It can get details wrong, so check the source.
Named in this story
NBER, Jensen Huang, arXiv, NVIDIA (NVDA)
A linked name opens that asset's sentiment page.