A descriptive model of agent-documentation interaction is derived as a two-lobed cycle rather than a linear journey, and it is shown that two widely assumed properties of"agent-friendly"documentation - actionability and verifiability - lack consistent behavioural support.
Abstract
Technical documentation is written for human developers, but an increasing share of software changes is now authored by autonomous coding agents. Which documents they consult, when, and what follows remain unknown. We conduct a behaviour-grounded study of agent-documentation interaction across two public datasets: 557 agentic coding sessions from SWE-chat, yielding 94,813 development events including 3,033 documentation interactions; and 33,097 agentic pull requests from AIDev, with 690,260 classified file-level change records. Four findings challenge current documentation practice. First, agents'documentation work is dominated by agent-facing artefacts: instruction files and working notes account for 60.5% of all documentation interactions, versus 10.6% for classical technical documentation and 1.3% for API references. Second, the link between consultation and code editing is unresolved: the adjacent transition probability is 0.002 and the unadjusted three-event lift 1.05, whereas a stage-adjusted model places it above unity (OR 1.33 [1.09, 1.62]); documentation creation is elevated unadjusted (lift 1.67) but its adjusted interval includes unity. Third, no explicit documentation-based validation sequence was observed, and consultation is associated with less immediate testing (lift 0.23, cluster CI 0.08-0.45; adjusted OR 0.39 [0.25, 0.60]). Fourth, consultation is self-initiated (70.2%) far more often than failure-driven (7.5%), and documentation trails code: among multi-commit pull requests changing both, code is touched first 4.7x more often. From these traces we derive a descriptive model of agent-documentation interaction as a two-lobed cycle rather than a linear journey, and show that two widely assumed properties of"agent-friendly"documentation - actionability and verifiability - lack consistent behavioural support. We release our pipeline, coding scheme, and event-level data.
Drug use is an ancient practice, but its associated disorders represent a contemporary public health challenge. This study investigates the impact of proximal processes in childhood/adolescence and adulthood on substance use, focusing on the role of Therapeutic Communities (TCs). Using a qualitative methodology, 19 residents of TCs in the state of Rio de Janeiro were interviewed. Instruments included a screening test (ASSIST), a sociodemographic inventory, and semi-structured interviews. Content analysis of the interviews was supported by the Requalify.ai software, which proved to be an efficient tool for categorizing and visualizing qualitative data. Results indicate that factors such as dysfunctional family environments, violence, and early onset of consumption, often mediated by peer influence, are determining risk factors. On the other hand, peer social support within TCs emerges as a crucial protective factor, associated with positive changes reported by participants. The sample revealed an overrepresentation of Black and Brown individuals, highlighting the racial dimension in the history of drug use in Brazil. The study concludes that proximal relationships are decisive in both the etiology and recovery of substance use disorders, and that TCs, although controversial, can offer a supportive environment that favors change, especially through peer support and cohabitation.
Marceli de Souza Rosa-Pereira, L. Pessoa· Lumen et Virtus· 0 citations
An audit-and-placebo protocol is proposed that separates verifier artifacts, interaction scaffolding, and grounded feedback credit in evaluations of self-evolving test generators in evaluations of self-evolving test generators.
Yunhao Liang, Chengguang Gan, Ruixuan Ying et al.· 0 citations
A reverse engineering technique is presented for extracting a behavioural model of a deployed chatbot through the automated exploration of its functionality via Large Language Models through the automated exploration of its functionality via Large Language Models.
Iván Sotillo del Horno, Alejandro del Pozzo, Esther Guerra et al.· SN Computer Science· 0 citations
This prototype MRG image translocation software was helpful to 69% of patients with binocular diplopia, but limited by large angle strabismus because of the limited instrument field of view.
Edsel B Ing, Kevin Sha, Sarosh Dandoti et al.· Journal of neuro-ophthalmolo...· 0 citations
A diagnostic support system based on a unified web platform that classifies patients according to the risks of developing three diseases based on regularly collected clinical or audio data using classical supervised learning algorithms is presented.
Vedamurthy D R, Dr. Anup Ritti, A. Bibi et al.· International Journal for Re...· 0 citations
Telecommunications base-station towers support services that are indispensable to crisis communication, emergency response, public administration and economic continuity. Their dispersed location, height and lattice geometry make conventional close-access inspection hazardous, time-consuming and difficult to document consistently. This study evaluates the usefulness and limitations of uncrewed aircraft systems for inspecting steel lattice towers that form part of telecommunications critical infrastructure. A retrospective field study was conducted using records from inspections performed in 2022–2023 at 50 towers in three Polish provinces. More than 150 sorties were completed with a DJI Matrice 30T platform using optical zoom, thermal imaging, a laser rangefinder and georeferenced mission telemetry. Manual, column and spiral flight patterns were compared, while image sets were processed with photogrammetric software to support three-dimensional reconstruction and spatial localisation of defects. The observations show that uncrewed inspection is particularly effective for documenting coating damage, local corrosion, fastener deterioration, connection anomalies, snow or ice accretion and equipment requiring thermographic follow-up. Column flights generated the most detailed evidence but also the greatest processing burden; spiral flights were faster and better suited to screening and three-dimensional modelling. The study also identified important constraints: wind, battery endurance, vegetation, airspace restrictions, electromagnetic and navigation disturbances, and the need to protect sensitive infrastructure data. Uncrewed aircraft should therefore be treated as a risk-based screening and documentation tool integrated with engineering judgement and targeted close-contact or non-destructive testing, rather than as an unconditional replacement for statutory structural inspection.
Rafał Parczewski, Tomasz Balcerzak, S. Żurawski· Zeszyty Naukowe SGSP· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduAug 17, 2026
A USAF cadet and a Lincoln Laboratory researcher found AI chatbots can help nontechnical service members produce viable software applications for their unique problems.
