Self-reported ai productivity keeps failing the telemetry test. an icse 2026 study of developer…
self-reported ai productivity keeps failing the telemetry test. an icse 2026 study of developer logs found 82.3% of developers feel more productive while typed characters grew for ai users and non-users alike, with no significant change in code quality.
the speedup lives in the survey, not the logs.
Context
The preprint Evolving with AI: A Longitudinal Analysis of Developer Logs (arXiv 2601.10258), in the ICSE 2026 research track and run with JetBrains, uses two years of IDE telemetry from 800 devices, 400 AI users who used JetBrains AI Assistant every month from April to October 2024 and 400 who never did, plus a survey of 62 professionals. In the survey 82.3 percent said AI coding tools slightly or significantly increased their productivity. In the telemetry, typed characters rose over time for both groups, both trends statistically significant, by 75 characters a month for non-users and 587 a month for AI users. The paper reports AI users produce substantially more code but also delete significantly more. For quality, 48.4 percent in the survey said it increased, while telemetry used debugging sessions as a proxy and found no significant change for AI users.
The 82.3 percent is a survey of 62 people who identified as AI-tool users, not a random sample. Both groups grew in typed characters, but AI users grew about 7.8 times faster, so alike understates the paper's finding of greater acceleration in active code authoring. Typed characters is a proxy for productivity and debugging sessions is a proxy for quality, so no significant change is not a quality verdict. The study was run with JetBrains, which sells AI in its IDE, using its own telemetry, and the survey and the logs are different populations. That the speedup lives in the survey, not the logs, is the author's reading; the logs do not measure speed or output quality.
Related work
- Earlier note on the same study ↗Same ICSE paper.
- Earlier note on the same study, first read ↗Same 800 devices.
Watch next
- Replication on non-JetBrains data and a direct task-time measurement.
Sources
- Evolving with AI: A Longitudinal Analysis of Developer Logs (arXiv 2601.10258)arxiv.org
- ICSE 2026 research track pageconf.researchr.org
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 20 September 2026 at 18:16 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →