WWDC 2025: iOS 26, Liquid Glass, and a Q3 Re-Prioritisation List

Apple's WWDC 2025 keynote ran on June 9, 2025, and it dropped three things that change an iOS roadmap: a year-based rename to iOS 26, a system-wide redesign called Liquid Glass, and a Foundation Models framework that puts an on-device LLM behind a few lines of Swift. The morning after, three Indian app teams we work with asked the same question — "what do we drop from Q3 to absorb this?" This post is the re-prioritisation list we actually handed them, with the items ranked by how badly they bite if you ignore them. ## TL;DR — What changed and what to do about it Three changes matter. First, Liquid Glass restyles every standard control, so a custom-tinted navigation bar or a hand-rolled tab bar will look broken on iOS 26 — audit your custom UI. Second, the Foundation Models framework gives you a free, private, on-device model for summarisation and extraction — move cheap LLM calls off your API bill. Third, Live Translation lands at the OS level, which changes the build-vs-skip math for in-app translation features. Test all three against the beta before the September release window closes. ## Why this matters now (and the dates that bind you) WWDC 2025 keynoted on June 9, 2025, with iOS 26, iPadOS 26, and macOS Tahoe 26 slated as free updates in fall 2025 ([Engadget's full recap](https://www.engadget.com/big-tech/wwdc-2025-ios-26-new-liquid-glass-design-and-everything-else-apple-announced-171718769.html)). That hands you a roughly 12-week beta window. The trap: teams treat the beta as "test in August." By August your customers are already on the public beta filing one-star reviews about a broken date picker. The work has to start the week the first developer beta drops, not the week before public release. We watched this exact pattern in 2024 — a Pune fintech app shipped a custom segmented control that rendered as a grey blob on the iOS 18 beta, and the bug report queue did the rest. The fix took four hours; the panic took four days. Front-load the audit. There's a second reason the timing is tight: App Store review behaviour shifts during a major OS transition. Reviewers test on the new OS, and a layout that clips or a control that's unreadable over the new material can draw a rejection or a metadata note you didn't expect. Submitting a routine update in August on an app you haven't tested against the beta is how a two-line bug fix sits in review for an extra week. Knowing your app already renders cleanly on iOS 26 removes that risk before you queue any release. ## The 5-item Q3 re-prioritisation list (ranked by bite) Here is the ranked list. Each item names the symptom you'll hit if you skip it, the rough effort, and the verification step. ## How much does the Foundation Models framework save you? The Foundation Models framework exposes Apple's on-device model to any app through Swift, at zero token cost and with no network round-trip ([CBS News confirmed the framework launch](https://www.cbsnews.com/news/apple-wwdc-2025-time-june-9-what-to-expect/)). For a class of tasks you're currently paying a cloud API for, that's a direct line-item cut. The catch: it only fits on-device-appropriate work — short summaries, tagging, extraction, simple classification. It will not replace a frontier model for long-context reasoning. We ran the same off-device-to-on-device pattern when cost-engineering our in-house voice app. On TalkDrill, our in-house English-speaking app, we route cheap classification tasks to the smallest model that clears a quality floor and reserve the frontier model for the hard ones. The Foundation Models framework is the same routing decision, except the cheap tier is now free and on-device. ## The DIY walkthrough: a 5-day iOS 26 readiness sprint This is the sprint we run. Each step has a verification so you know it's actually done, not just attempted. ## When NOT to chase every WWDC feature Not every announcement deserves Q3 engineering time. Skip on-device LLM work if your app's AI tasks are all long-context reasoning — the on-device model is too small, and you'll spend a week proving what the docs already say. Skip a full Liquid Glass redesign if you already use standard system controls everywhere; you'll get most of the new look for free on recompile. And skip building any feature that depends on a beta-only API for a launch that has to ship before the public release — beta APIs change, and you'll eat the rework. The honest version: for a team using mostly standard UIKit or SwiftUI components, iOS 26 readiness is a 3–4 day audit, not a quarter-long rebuild. The teams that suffer are the ones who reinvented system controls two years ago and now own that maintenance forever.

• iOS 26 developer beta installed on a non-primary test device
• Before-screenshots captured for every screen
• Custom controls audited: keep / replace / restyle decided
• One Foundation Models on-device call shipped and verified offline
• Live Translation build-vs-defer decision recorded
• WCAG AA contrast re-checked over busy backgrounds, both transparency modes
• Layout regression run on smallest and largest devices
• App Intents added for top 3 user flows (if previously skipped)
• Beta-dependent features flagged as not-for-this-release

## A real example: how we re-planned a 7-week mobile build We were mid-build on a HIPAA-aware clinic app when an OS redesign landed in a prior cycle, and the lesson carried straight into our WWDC 2025 advice. On that project — documented in our write-up of building a clinic mobile app in 7 weeks — we'd used mostly standard navigation, so the redesign cost us a single day of contrast fixes rather than a re-skin. The team that had hardcoded a tinted toolbar two screens over spent a week. As Khushi, our UI/UX designer, frames it on every iOS engagement: use the system control unless you can name the specific reason you can't. That one rule is what makes a WWDC redesign a 3-day audit instead of a re-platform. Our mobile development team bakes it into the design review before a single screen ships. The same discipline shows up in our voice work — see the engineering behind TalkDrill, where we kept the UI standard so OS shifts stay cheap. ## How Liquid Glass changes design-system decisions Liquid Glass isn't only an engineering concern — it resets a few design-system assumptions too. The translucent material means your brand colour now sits behind a frosted layer, so the exact hex you picked in 2023 reads differently over a moving background. Solid-fill buttons that looked confident on iOS 18 can feel heavy against the new lighter chrome. And depth now comes from the material's blur and layering rather than drop shadows, so a shadow-heavy design language starts to look dated next to native controls. The practical move is to treat your design tokens as a review item, not a constant. Re-check the three that shift most: primary-button fill, surface elevation, and text-on-translucent contrast. For most apps this is a half-day in the design file plus a build to confirm, not a rebrand. The teams that get burned are the ones that treat the OS look as fixed and let their app drift further from native every year — by iOS 28 they're maintaining a visual language nobody else uses. For the 2026 angle on Apple's roadmap — the Gemini-powered Siri and App Intents as a distribution channel — see our companion piece, WWDC 2026 Siri teaser for Indian app devs. And for the founder's first-person read on Apple's AI direction, Vivek Singh's blog covers the platform-strategy side. ## FAQ ### When was the WWDC 2025 keynote and when does iOS 26 ship? The WWDC 2025 keynote ran on June 9, 2025. iOS 26, iPadOS 26, and macOS Tahoe 26 were announced as free updates slated for fall 2025, typically September. That gives developers roughly a 12-week beta window to test and fix before the public release. ### What is Liquid Glass and will it break my app? Liquid Glass is Apple's system-wide redesign that gives most UI elements a translucent, glass-like material. It will not break apps using standard controls — those get the new look on recompile. It can break apps with custom-tinted bars, hardcoded blur, or hand-rolled controls, which may render incorrectly. ### What is the Foundation Models framework? The Foundation Models framework lets any app call Apple's on-device LLM through Swift, at zero token cost and with no network round-trip. It suits short summarisation, tagging, extraction, and simple classification. It is not a replacement for a frontier cloud model on long-context reasoning tasks. ### Should I remove my in-app translation now that Live Translation exists? Not automatically. Live Translation works inside Messages, FaceTime, and Phone at the OS level. If your translation feature lives in your own UI or needs domain-specific terms, the OS feature may not cover it. Decide per use case: defer to OS where it overlaps, keep where you differentiate. ### How long does an iOS 26 readiness audit take? For a team using mostly standard SwiftUI or UIKit components, the audit is roughly 3–5 days: a UI pass, one Foundation Models migration, a contrast recheck, and a layout regression sweep. Teams with heavily customised controls should budget more for the replace-or-restyle work. ### Do I need to rewrite anything to support Liquid Glass? Usually no full rewrite. Recompiling against the new SDK adopts the material on standard components. The work is auditing custom UI, fixing contrast over translucent surfaces, and catching small layout shifts — not re-architecting your app. ### Is the on-device model good enough for a chatbot? For short, structured tasks inside a constrained domain, often yes. For an open-ended support chatbot that needs long context, retrieval, and nuanced reasoning, no — route those to a cloud model. The right pattern is task-based routing: cheap and on-device for the easy 80%, cloud for the hard 20%.