Honolulu's Department of Planning and Permitting has opened a priority-review fast track: residential plans that clear CivCheck — an AI tool that flags missing documents and common code issues before you file — route straight into prescreening. A pilot batch moved through prescreening in about 32.5 days versus roughly 70 for projects that skipped it, and the city credits CivCheck with cutting residential code-review time by roughly 70%. It's free, residential-only for now (commercial types promised in FY27), and slated to become mandatory for qualifying residential work later this year. Contractors told Civil Beat the simple projects already come back cleaner; the real test is complex commercial work. The notable part isn't another permitting startup — it's the authority having jurisdiction itself running the check, ahead of human plan review. Hoodline →

A targeted scan of the early-design frontier — programming, massing, feasibility-to-design — turned up only vendor launches and "top-10 AI tools" listicles, nothing clearing the bar of independent reporting or a named outcome. Worth noting where movement in this phase actually is: not in a new massing app, but upstream, in the developer-side feasibility tools above that increasingly hand an architect a program and a site plan before the first design meeting.

A systematic review in Buildings (21 empirical studies, 2019–2026) draws a sharp line through the "AI clash detection" category: detection is now operationally mature, but the stages that matter for coordination — resolving clashes and preventing them upstream — still lean on manual expert judgment. Systems marketed as "autonomous," the authors find, have only ever been validated in simulation, never on live projects, and most produce black-box outputs that "cannot meet the accountability requirements of professional coordination practice." For the architect or MEP engineer weighing a clash-detection add-on, that's the useful frame: the tool can tell you where models collide; deciding whose model moves, and why, is still yours. Buildings (MDPI) →

McLaren Construction is deploying autonomous robot dogs on UK sites — Boston Dynamics Spot units running California startup FieldAI's "Field Foundation Models," which fuse data-driven AI with physics-based reasoning so a robot can navigate stairs, doors and a changing site with no prior map. The first missions are squarely documentation: 360° imagery, point clouds, progress verification, model-to-site deviation, and safety and QA patrols, tested on the Passivhaus refurbishment of the LSE's 35 Lincoln's Inn Fields. FieldAI — backed by Bezos Expeditions, Nvidia, Gates and Samsung to the tune of $405M — says it's already on "hundreds of sites." Monday's robotics news was lab scaling curves; this is a named GC on real jobs. Closer to home, Birmingham's Robins & Morton and Auburn's CARV Lab have run Spot on the same premise — the new part is the maplessness. Global Construction Review →

Dallas–Fort Worth laid out how it's rebuilding operations and maintenance for a climb to 100 million passengers and a $12B expansion. It fused work orders, sensor feeds, inspection notes and OEM lifespans into a single "Asset Health 360" index, with a digital twin that auto-generates condition-based work orders and an AI scheduler matching jobs to technicians in real time; the target is a 20–25% cut in maintenance cost per square foot by 2030. Two details matter more than the tech: the team used AI to accelerate the unglamorous data-cleansing first, then ran the program "outcomes first" with the frontline in from week one — landing above 90% technician adoption. The digital twin was demoted from north star to "one layer." The operating model did the work. International Airport Review →

Gemini 3.5 Pro — reported for release around July 17 with a 2M-token context window, it's the last major Western flagship still pending. Double the context matters for reasoning across a whole project's document set. Odaily →

FieldAI's next step — whether the robots graduate from "monitoring missions" to physical work (site logistics, dexterous manipulation, multi-robot coordination) is McLaren's stated path. Reliability on messy, unstructured tasks is the wall that has stopped construction robotics before. Construction News →