Sixty-two percent of San Francisco-based AI candidates surfaced in ENTRA's Q3 2026 Recruiter Survey (n=1,140 active AI job-seekers, Bay Area metro, July–August 2026) are actively re-querying for distributed or remote-first roles — up from 41 percent in the same cohort one year earlier. The proximate cause: Google began incorporating badge-scan data into performance reviews at its Mountain View and Sunnyvale campuses in Q2 2026, moving a three-day hybrid expectation toward near-five-day attendance for engineering and research teams. Meta pushed Instagram to a full five-day office requirement in February 2026 and extended office-first expectations to senior ICs across other product divisions through the summer. Salesforce, which Marc Benioff declared a "remote-first" company in 2022, now requires sales and technical staff to attend four to five days per week from its San Francisco headquarters at 415 Mission Street. The Bay Area's second RTO wave has arrived, and the 62 percent signal shows exactly where the pressure is landing.
The Mandate Wave
Google: Badge Data Meets Performance Review
Google's move from three-day hybrid to five-day practical enforcement did not arrive via a single all-hands memo. It arrived via infrastructure. Google began using badge-scan records to track attendance against stated hybrid policies in 2023; by Q2 2026, teams at the Mountain View campus were told that scan data — not self-reported compliance — would factor into their performance review inputs. Multiple Google Research and Google DeepMind engineers described the shift to The Information in Q2 2026 as "soft five-day" enforcement: no formal policy had changed in writing, but non-attendance was now formally visible in the review record and being raised in 1:1 conversations with managers. For an organization that employs approximately 12,000 AI-focused engineers across its Mountain View, Sunnyvale, and San Francisco offices (ENTRA Job Signal Index headcount estimates, Q2 2026), the badge-to-review linkage is a structural change in how attendance accountability works — regardless of what the policy document says.
Meta: Instagram Sets the Template
Meta's Instagram division moved to five-day in-office attendance effective February 2, 2026. Instagram chief Adam Mosseri outlined the change in an internal memo titled "Building a Winning Culture in 2026," stating that "we are more creative and collaborative when we are together in-person" — a line that subsequently appeared in press coverage of the mandate. The broader Meta policy for most divisions remained three days on paper through mid-year, but ENTRA Q3 2026 Recruiter Survey respondents currently employed at Meta described the practical expectation for senior ICs at the IC5 and IC6 levels as four to five days, with remote exceptions requiring explicit director-level or above sign-off. The Instagram pilot, which Meta management has not publicly disavowed, functions as a template that can extend to other divisions without a formal policy announcement.
Salesforce: Remote-First in Reverse
Salesforce CEO Marc Benioff's 2022 declaration that the office was "dead" has not survived the company's 2025–2026 performance cycle. Beginning October 2024, Salesforce required sales teams and select technical staff — covering data center engineering, onsite support, and workplace services — to return four to five days per week. By Q1 2026, the "office-flex" designation covering marketing, legal, and product teams set a three-day minimum. The reversal carries brand cost that goes beyond HR policy: Salesforce's remote-first stance had been a centerpiece of its recruiting collateral for two full years. Glassdoor trend data for Q1 2026 showed "work from home" moving from the top-five pros list to the top-five cons list for Salesforce — the first time that marker had flipped negative since 2020.
The Talent Response
Flight to Distributed-Native AI Infrastructure
The departure pattern that has emerged from ENTRA Job Signal Index Q3 2026 tracking — monitoring Bay Area AI candidates who updated LinkedIn location preferences or opened new applications between July 1 and August 31, 2026 — is not diffuse. It concentrates on a specific set of employers: distributed-native AI infrastructure companies that have never operated an office-first posture and are not likely to adopt one.
Weights & Biases, which runs a globally distributed team of approximately 260 and has operated remote-first since its 2018 founding, saw inbound candidate volume from San Francisco-headquartered applicants increase 38 percent quarter over quarter between Q2 and Q3 2026, per ENTRA Job Signal Index data. The company's ML Engineer and MLOps roles drew particularly from engineers at Google and Meta listing RTO policy as the primary trigger in ENTRA's candidate intake survey. Modal, the cloud-native AI infrastructure platform that maintains a distributed engineering team despite its San Francisco origins, and Replicate, which runs model inference infrastructure with a remote-first engineering org, are absorbing a parallel cohort: ML engineers with deployment and inference backgrounds who have no appetite for a five-day commute to Santa Clara or Menlo Park. Together AI, whose distributed research and engineering team spans the US without an office attendance requirement, added 22 open roles in the Bay Area catchment area in Q3 2026 — a 31 percent increase from Q2 — per ENTRA Job Signal Index.
Geographic Redistribution: East Bay and South Bay Diverge
Not all movement is toward fully remote employment. A portion of the Bay Area AI talent pool is executing an analog arbitrage: relocating from San Francisco to the East Bay — Oakland, Berkeley, Emeryville — where commute time to Mountain View, Sunnyvale, or San Jose offices rises from 45 minutes to 90 minutes each way. That additional commute time raises the friction of a five-day week to a threshold that converts ambivalent employees into confirmed job-seekers. ENTRA Q3 2026 Recruiter Survey respondents in the East Bay listing Google or Meta as their current employer were 2.1x more likely to have opened a new job search in the previous 90 days than comparable respondents located in San Francisco proper.
The South Bay tells a different story. Apple engineers in Cupertino and NVIDIA engineers in Santa Clara — both companies maintaining comparatively durable hybrid cultures — reported re-evaluation rates of 38 percent and 34 percent respectively in the ENTRA Q3 2026 survey, materially below the 62 percent Bay Area average. The geographic distribution of re-evaluation rates maps almost exactly to the geographic distribution of five-day enforcement severity.
The ENTRA Q3 2026 Data Signal
The 62 percent figure from ENTRA's Q3 2026 Recruiter Survey (n=1,140, Bay Area metro, conducted July–August 2026) breaks down by enforcement context. Among AI candidates at companies that have implemented five-day or near-five-day office requirements — Google, Meta/Instagram, Salesforce, Amazon — the share re-querying for distributed-first roles rises to 71 percent. Among candidates at companies with three-day hybrid policies that have not announced further tightening, the re-evaluation rate sits at 44 percent: elevated from the historical baseline but not in the same active-search cohort.
Role type tracks the same pattern. ML Engineers and Research Engineers report the highest re-evaluation rates at 68 percent — a job family where the distributed-native alternatives (Weights & Biases, Modal, Replicate, Together AI) are directly competitive in both scope and compensation, and where the remote-work trade-off has largely collapsed. Data Scientists and AI Product Managers report lower re-evaluation rates at 54 percent, reflecting fewer distributed-native alternatives at equivalent total comp.
ENTRA Recruiter Survey respondents were asked directly: "If your current employer enforced five-day in-office attendance, what would you do?" Forty-four percent said they would seek a new role within 90 days. Thirty-one percent said within six months. Eighteen percent said they would comply. Seven percent were undecided. For a Bay Area AI market employing approximately 87,000 workers in AI-relevant roles (BLS Occupational Employment projections, July 2026 update), the 44 percent immediate-search cohort alone represents a potential active candidate population of approximately 38,000 — a number that reshapes recruiter pipeline math at every distributed-native company in the space.
What Comes Next
The Q4 2026 signal to watch is whether Google formalizes badge-tracking enforcement into explicit policy language before year-end performance reviews close in November. If attendance data moves from a review input to a stated performance criterion — the step Google has so far stopped short of — expect the 62 percent re-evaluation figure to push toward the 71 percent range currently seen only in five-day enforcement cohorts. Meta's mid-year extension of office-first expectations to IC5 and IC6 levels will face a retention test when its annual comp review closes in December; if the exits that follow match the survey intent, Facebook, WhatsApp, and Messenger engineering teams become the next wave. Salesforce has the longest road back: reversing a remote-first talent brand takes multiple recruiting cycles to repair, and the distributed-native AI infrastructure sector it inadvertently energized is hiring into the gap right now.
Global Career PlatformFind AI talent. Find your next role.
Booking is hotels. · Airbnb is apartments. · ENTRA is global careers.