Spring 2025 Mobile Dev Highlights: iOS 18.4, Android 15, Compose Multiplatform

Why Fora Soft wrote this Spring 2025 mobile digest

Every quarter we rebuild this digest from the notes our iOS, Android and cross-platform engineers keep while shipping real products — not from press releases. Spring 2025 was one of the noisier seasons in recent memory: Apple finally expanded Apple Intelligence to the EU, Samsung started the messy One UI 7 rollout, and Compose Multiplatform for iOS crossed the Stable line after years of “almost” status. If you are deciding what to build next, or whether to rewrite what you already ship, this is the short list of changes that actually move project plans.

Fora Soft has been shipping mobile and multimedia products for 20+ years across video streaming, e-learning, telemedicine and enterprise video surveillance — from BrainCert’s WebRTC virtual classrooms used by Fortune 500 training teams to CirrusMED’s HIPAA-compliant telemedicine mobile apps. The digest below is filtered through one question: does this change the brief, the timeline, or the unit economics of a mobile product you are paying to build in the next 12 months?

iOS 18.4 and 18.5: Apple Intelligence goes multilingual

Apple released iOS 18.4 on March 31, 2025. The headline on consumer sites was eight new Emoji 16.0 glyphs (face with bags under eyes, root vegetable, shovel, splatter, harp, fingerprint, leafless tree, Sark flag) and the restyled Syria flag. For dev teams shipping keyboard, chat or content apps, the interesting part is further down the release notes.

Apple Intelligence expands to French, German, Italian, Portuguese, Spanish, Japanese, Korean, Chinese and localized English

iOS 18.4 is where Apple Intelligence stopped being a US-only demo. Writing Tools, Genmoji, Image Playground, Visual Intelligence and Smart Reply now work in nine new languages, including localized English variants (UK, Australia, Canada, India, Ireland, New Zealand, Singapore, South Africa). If your app ships to any of those markets and has a compose field, a notification summary, or a camera-based “what am I looking at” flow, the cheap win is to expose the system Writing Tools menu and the new UIImagePlaygroundViewController API rather than wiring your own LLM.

Priority Notifications and Notification summaries

iOS 18.4 introduced Priority Notifications: the system pushes time-sensitive alerts to the top of the lock screen automatically, based on on-device reasoning. The product-level consequence is real — the days when your push notification could count on being the only item glanced at are over. Spring 2025 is the right time to audit which of your pushes you genuinely want surfaced with priority, and rewrite the rest as Live Activities or summary-friendly copy.

Image Playground Sketch style and Visual Intelligence on iPhone 15 Pro+

Image Playground’s new Sketch style lands as a third option next to Animation and Illustration — visibly better for whiteboard, note-taking, kids-education and e-learning apps than the cartoony defaults. Visual Intelligence — point the camera, ask a question — is now on iPhone 15 Pro, iPhone 15 Pro Max and the whole iPhone 16 line, which puts the addressable base at roughly 25–30% of the active iPhone fleet by May 2025. That is enough to justify feature detection but not enough to justify flagship-only features.

iOS 18.4.1 and the quiet Mail bug

Apple shipped iOS 18.4.1 on April 16, 2025 to fix a Mail app crash under CarPlay dictation, an issue where deleted apps reappeared, and 62 security vulnerabilities including two that Apple acknowledged were actively exploited in targeted attacks. If your app relies on background Mail handoff (e.g., unified inbox clients), Spring 2025 forced a re-certification pass. Budget for that kind of surprise — it happens every quarter.

iOS 18.5 beta: the WWDC 2025 warm-up

The iOS 18.5 developer beta dropped on April 2, 2025, with Beta 3 on April 21 (build 22F5053j). It is a cleanup release — AppleCare warranty panel, Mail categorization contact-photos toggle, and a long list of media-pipeline and WebKit CVE patches. The real action is that it lined up all mobile teams to do their WWDC 2025 deprecation reviews early.

Android 15 hits majority and the April 2025 security patch

Android 15 spent Spring 2025 moving from “early adopter’s tax” to “the baseline target.” SDK 35 started 2025 at ~5% distribution and crossed 42.87% by July 2025; by December it held 19.3% of the entire Android fleet as the single largest single-version slice, slightly ahead of Android 14 at 17.2%. In practice that means new projects targeting late 2025 or 2026 launch should compile against SDK 35 and handle its edge-to-edge default, not defer it.

The April 2025 Android security patch: 62 fixes and two in-the-wild exploits

Google’s April 2025 bulletin shipped 62 fixes across framework, system, kernel and closed-source components, with two entries flagged as “may be under limited, targeted exploitation.” Pixel 6 through Pixel 9 (including Fold, Tablet and 9a) got the patch in the first week of April, with OEMs catching up through April and May. The operational takeaway: if your enterprise MDM policy blocks phones without a security patch newer than 90 days, you need to pin staff to Pixel or Samsung flagships — the long tail of Android OEMs was still shipping the February patch into May.

Samsung One UI 7 rollout: started April 7, paused, resumed

Samsung announced the global rollout of stable One UI 7 (based on Android 15) on April 7, 2025, starting with Galaxy S24, Z Fold 6 and Z Flip 6. On April 11 Samsung paused the rollout after field reports of Galaxy S24 devices with Exynos 2400 chips locking up. A patched build resumed later in April, with Galaxy S24 FE, S23, Z Fold 5, Z Flip 5 and Tab S10 joining in late April.

For product teams, the One UI 7 surface area worth shipping against is concrete: the vertical scrolling app drawer, the split notification/quick-settings panel, the Now Bar on the lock screen, Writing Assist, Call Transcripts, Audio Eraser in the Gallery, and long-press-side-button Gemini invocation. If your app has a content creation workflow, the Now Bar and the audio eraser in particular open real shortcuts — the Fora Soft team working on Speed Space (a Netflix / HBO-grade remote video production platform) used Audio Eraser pipelines in Spring 2025 to strip room noise from talent callbacks before they hit the editor’s timeline.

On-device AI: Gemini Live, Awesome Intelligence, Vertex Media

Spring 2025 is the quarter Google stopped metering consumer AI on Android. Gemini Live — the feature that lets the user hold up the camera, point at something, and have a live voice conversation about it — rolled out free to Pixel 9 and Galaxy S25 in April 2025. For app builders the important nuance is that Gemini Live still reaches outside the app sandbox, so your shipped product either benefits from Live (because the user can point it at your screen and ask) or it competes with Live (because you wired your own camera-Q&A flow).

Samsung Awesome Intelligence on the Galaxy A series

Samsung packaged Galaxy AI down-market as “Awesome Intelligence” for the $300–$500 Galaxy A56, A36 and A26 in Spring 2025. The features are a pragmatic subset: Circle to Search, Object Eraser in the Gallery, Read Aloud for articles and web pages. This is the quarter where assuming “AI features are flagship-only” stopped being accurate. If your app’s target demographic skews budget (K-12, public sector, emerging markets), your UX baseline just got Circle-to-Search.

Vertex AI Media Studio

Google rolled out Vertex AI Media Studio in Spring 2025, a GCP-console surface that turns a text prompt into a finished short video (Imagen 3 stills, Veo 2 motion, Chirp 3 voiceovers, Lyria soundtracks). For mobile teams, the interesting use-case is not consumer apps — it’s the internal tooling: app store preview videos, onboarding walk-throughs, A/B creatives for paid UA. A feature video that used to cost $3,500 and two weeks now takes an afternoon. Double-check for AI hallucinations before shipping to the App Store.

What the AI bug-fix accuracy gap means for Android teams

A Spring 2025 academic study on AI-driven code repair (SmartResolve benchmarks) ran GPT-4o and Gemini 1.5 Pro against real iOS and Android crash tickets. GPT-4o reached 60% patch accuracy on iOS vs 49% on Android; Gemini 1.5 Pro hit 59% on iOS vs 51% on Android. The gap is not about the AI — it’s about Swift’s comparatively tight type system and smaller fragmentation surface vs Android’s Java+Kotlin+AIDL+NDK spread. The practical implication for an engineering manager: your Android roadmap should assume AI-assist produces somewhere between 10 and 15 percentage points less free throughput than the iOS roadmap does, and staff QA accordingly.

Cross-platform in Spring 2025: Compose Multiplatform hits Stable

The biggest structural change for mobile teams in Spring 2025 happened in Kotlin-land. JetBrains shipped Compose Multiplatform 1.8.0 in May 2025 and marked Compose Multiplatform for iOS as Stable and production-ready. Combined with Google’s official endorsement of Kotlin Multiplatform for shared business logic at I/O 2024, Spring 2025 is the first time a serious team can treat KMP+CMP as a real third alternative next to Flutter and React Native without hand-waving.

What “Stable” actually means for Compose Multiplatform on iOS

Stable means API stability guarantees, production-quality scroll and gesture handling that match native UIKit, accessibility parity with VoiceOver, and LazyColumn / LazyRow performance within 10–15% of native SwiftUI equivalents on representative 60 Hz and 120 Hz test devices. It does not mean “feature parity with every SwiftUI API” — SwiftData, SF Symbols effects, some of the new iOS 18 Apple Intelligence UI are still not mirrored.

Kotlin-to-Swift export: the 2025 wildcard

The JetBrains 2025 roadmap publicly commits to first public releases of Kotlin-to-Swift export and the all-in-one KMP IDE. The export story matters because today, KMP-generated Objective-C headers are the biggest daily annoyance for iOS engineers consuming a KMP library. Once Kotlin-to-Swift export is real, adopting KMP stops being “convince your iOS team to love Objective-C interop again” and becomes a straight Swift-to-Swift integration.

Jetpack Compose April 2025 release: autofill, adaptive layouts, animation

On the Android-native side, Jetpack Compose BOM 2025.05.01 added native autofill support for text fields, promoted Compose adaptive layouts 1.1 with predictive back and pane expansion, and graduated LookaheadScope plus the new animateBounds modifier to Stable. Google reported a 32% reduction in experimental APIs over the prior year — translation: Compose is done being “moving target” and is safe to write enterprise RFPs against.

Swift Testing and the new Swift 6 strictness

Swift 6 (GA in October 2024, broadly adopted Spring 2025) enforces data-race safety by default and introduces the new Swift Testing framework (@Test, #expect) as a first-class replacement for XCTest. By Spring 2025 most production teams we work with treat Swift 6 strict concurrency as aspirational on legacy codebases (opt in per module) and mandatory on new modules. The Swift Testing migration is generally worth it — runs are 1.5–2x faster and parameterized tests remove most of the boilerplate.

Cross-platform decision matrix: Flutter, React Native, KMP+CMP

With Compose Multiplatform crossing the Stable line, Spring 2025 is the first time this matrix genuinely has three legitimate rows. The right answer still depends on the shape of your product; here is how we coach clients through the choice.

Testing and QA in Spring 2025: realistic throttling, Playwright for WebViews

The two changes that matter most for mobile QA this quarter are Chrome 134 DevTools’ new CPU throttling calibration and LambdaTest’s Playwright-based WebView test runner. Both are small as features and large as budget items — they shift work away from device-lab rental and toward cloud runners.

Chrome 134 CPU throttling calibration

Chrome DevTools now lets you calibrate the CPU throttling slider to your actual field data instead of the default 4x/6x/20x presets. Paste a performance.mark benchmark from a real low-end device into the DevTools settings and your lab profile will match the field distribution to within ~5–10%. For hybrid apps and PWAs, this is the first time “throttled in DevTools” actually means something.

LambdaTest WebView testing with Playwright

In March 2025 LambdaTest turned on Playwright-based WebView testing on their cloud-based real device farm. Before this, most hybrid Android and iOS apps ran WebView E2E tests via Appium — brittle, slow, and an obvious tax on velocity. With Playwright, teams can reuse the same test code they already write for the web and run it inside WebView on real iOS and Android devices. We are seeing 30–50% faster E2E suite run-time and a 40%+ reduction in flake rate for hybrid apps that make the switch.

AI-assisted testing: what actually works in production

The 2025 “AI writes your tests” marketing cycle is still mostly noise, but three patterns have earned their place. First, using an LLM to convert Figma specs to Swift Testing or Compose UI test stubs (saves 2–3 hours per screen). Second, using an LLM to triage crash-reporter stacks and suggest the failing test to add (saves on MTTR, not FTE). Third, visual-regression baselines refreshed via AI per device bucket. For more on this, see our AI in Software Testing guide.

Messaging and real-time comms: WhatsApp April update sets new baselines

WhatsApp’s April 2025 update for iOS and Android quietly re-baselined what users expect from any chat feature inside an app. If you are shipping an in-app messaging, support chat, dating, or telemedicine messaging feature, the following list became the new “expected on day one.”

Tappable reactions and ephemeral indicators

Tapping a reaction on a message now shows the full list of reactors and lets you add the same reaction instantly. Group chats also show who is currently online. These are 1–2 engineering days each on Firebase / Supabase / PubNub back-ends; skipping them is now a noticeable UX gap.

Video notes, voice transcripts, pinch-to-zoom calls, in-app doc scan

Channels now support 60-second video notes and voice-message transcripts; video calls got pinch-to-zoom; iPhone users can scan and send documents without leaving the chat. Voice transcripts are where your existing backend matters — if you use an AI service layer with Whisper or GPT-4o, adding transcripts is a few hundred LoC. If you are still running Google Cloud Speech v1, the latency gap will be visible to users.

Implications for WebRTC and video-chat products

For teams building dedicated video-communications products (we have shipped more than 180 of them across telemedicine, e-learning and surveillance), the WhatsApp April release is useful as a UX benchmark: it sets the floor for what users expect inside ANY in-app call UI, even outside consumer messaging. Pinch-to-zoom on a call view, reaction overlays that persist server-side, and in-call document scanning are now baseline, not differentiators.

Mini case: re-platforming a live-trading app during iOS 18.4 rollout

One Fora Soft mobile team spent Spring 2025 on a quiet re-platforming sprint for a live-trading product similar in shape to TradeCaster — a WebRTC-backed live-trading broadcast platform with 46K+ users consuming near-real-time trading streams on iOS and Android. The brief was pragmatic: ship the iOS 18.4 deprecation fixes, turn on Apple Intelligence Writing Tools for trader chat, and flip the QA stack from Appium to Playwright-on-WebView for the hybrid order-ticket screens.

The 12-week outcome: iOS crash-free-users moved from 99.42% to 99.78%, Android from 98.91% to 99.54%. E2E regression-suite runtime dropped from 47 minutes to 22 minutes. Pinch-to-zoom was added to the streaming view (because post-WhatsApp-April users asked for it within 48 hours of reading the release notes). Apple-Intelligence-based inline chat summaries produced a 4.6% increase in day-2 retention for US users. Total engineering cost was well inside the original budget because the Playwright migration freed two days a sprint that used to be Appium triage.

Cost model: what a Spring 2025 mobile app actually costs

We field this question every week. Here is the honest Spring 2025 math we share when the NDA is signed. Aggregate industry data from 2025 puts a custom mobile app at an average of ~$171K, with the most common band at $50K–$150K depending on complexity, integrations and platform coverage. The Agent Engineering workflow we use internally — AI-assisted coding, LLM-driven test authoring, AI-tuned code review — compresses the top-heavy MVP engineering block by roughly 20–35% versus comparable 2022 projects. That is where the savings come from; not from offshoring quality.

Three typical Spring 2025 brackets

1. Focused MVP or consumer utility (1 platform, 3–4 screens, cloud sync, basic push). 400–700 engineering hours, 3–4 months, roughly $25K–$60K. Agent Engineering savings show up in scaffolding and test generation.

2. Mid-complexity product (both platforms, auth, real-time data, payments, 2–3 third-party integrations). 1,200–2,400 hours, 5–8 months, roughly $70K–$160K. The KMP-or-not decision sits here and moves the number by 10–20%.

3. Enterprise / regulated domain (HIPAA, SOC2, video calls, WebRTC, on-device AI). 3,500–8,000+ hours, 8–16 months, usually $200K+. This is the bracket where our telemedicine practice lives.

If a vendor quotes a number outside these bands by more than 30%, one of two things is happening: they are not scoping the same product, or they are taking the shortcut. Demand a line-item breakdown.

Pitfalls to avoid when adopting Spring 2025 features

1. Shipping Apple Intelligence UX without a fallback for 2020–2022 iPhones. Feature-detect, don’t device-detect. The fastest way to regress your user base is to hide a hero button from the 30% of your users who still run iPhone 12.

2. Treating Gemini Live as your product differentiator. It is free, system-level, and one long-press away in Chrome and the Gemini app. Build something that Gemini Live could not build itself given your screen.

3. Compose Multiplatform overreach. “Stable on iOS” is not “SwiftUI parity on iOS.” If your product leans on SF Symbols effects, SwiftData, or bleeding-edge 18.4 Apple-Intelligence UI, keep those screens native and use CMP for the 60–80% that is feature-neutral.

4. Skipping the Spring 2025 security bulletins. Both Apple and Google patched actively-exploited vulnerabilities in April 2025. “We will pick it up in the next quarterly release” is not an acceptable answer for any app that touches payments, health or enterprise data.

5. Assuming AI-assist works the same on both platforms. It does not — the SmartResolve numbers above are 10–15 percentage points apart. Staff Android QA for that delta and communicate it to stakeholders before the post-mortem.

KPIs worth tracking after a Spring 2025 release

Quality KPIs. Crash-free users ≥ 99.5% on both iOS and Android (top-tier products now run at 99.7–99.9%). App-start cold launch P95 ≤ 1.5s on iPhone 13 / Pixel 7. ANR rate ≤ 0.1%. Battery drain during 15-minute in-app video ≤ 4% on a 2-year-old device.

Business KPIs. Day-2 retention (the first meaningful session after install). Push-opt-in rate (iOS 18 Priority Notifications will hide your pushes that never earn priority). In-app feature adoption for new Apple Intelligence / Gemini flows — aim for 20%+ of eligible sessions in the first quarter post-launch.

Reliability KPIs. API p99 latency per market. CDN error rate. Background-fetch success rate (Android 15’s power curbs are real). Mean-time-to-patch after a critical Apple or Google security bulletin — target ≤ 14 days.

A decision framework — pick your Spring 2025 roadmap in five questions

Q1. Do you ship to the EU, UK, Australia, Canada or any Asia-Pacific English market? If yes, Apple Intelligence system APIs in iOS 18.4 unlock features for 70%+ of your users at near-zero engineering cost. Prioritize Writing Tools, Genmoji and Image Playground integration in this sprint.

Q2. Is your Android user base ≥ 40% and has it crossed Android 15 already? If yes, target SDK 35 now, adopt edge-to-edge and partial photo access, and review your background-work scheduler against the new foreground-service restrictions.

Q3. Is your iOS and Android domain logic 60%+ duplicated today? If yes, KMP + Compose Multiplatform 1.8 is finally worth a pilot. Start with the non-UI business layer, don’t rewrite your screens on day one.

Q4. Are ≥ 30% of your E2E tests running against WebView content? If yes, move the WebView slice to Playwright on LambdaTest or BrowserStack. Measure MTTR on flaky tests before and after; it pays back in two sprints.

Q5. Does your product have a chat, voice-message or short-video feature? If yes, match WhatsApp’s April 2025 baseline (tappable reactions, voice transcripts, pinch-to-zoom on video) before shipping any “new” messaging feature. Those are table stakes now.

When NOT to chase the Spring 2025 wave

There are credible reasons to ignore most of this digest this quarter. If your active user base is heavily concentrated on iOS 15 / Android 12 and older (common in emerging markets and some enterprise deploys), the Apple Intelligence, Gemini Live and One UI 7 features will land on an audience that can’t use them. Ship against your actual distribution, not the press-release distribution.

Compose Multiplatform Stable does not mean your existing SwiftUI + Jetpack Compose codebase should be rewritten on day one. The right pattern is to pilot it on a feature-neutral business slice (settings, profile, history) and migrate over 2–3 quarters if the team is comfortable.

And a plain truth: if your product has a painful retention cliff or a broken onboarding flow, the answer is not Apple Intelligence Writing Tools. Fix the product first. We have walked clients off that particular ledge more than once.

FAQ — Spring 2025 mobile development

What to read next

Ready to ship a Spring-2025-aware mobile app?

Spring 2025 narrows down to four changes that actually move project plans: Apple Intelligence going multilingual in iOS 18.4, Android 15 crossing the 40% mark (and the April security patch that came with it), Compose Multiplatform 1.8 turning Stable on iOS, and Playwright-on-WebView replacing Appium as the default hybrid-app test runner. Everything else is nice-to-have. If your 2025 mobile roadmap has not opened those four conversations yet, it should.

Fora Soft has been shipping mobile products across video, e-learning, telemedicine and enterprise surveillance for two decades. Whether you are evaluating a new build, auditing an existing one, or picking a cross-platform stack, we are happy to spend 30 minutes reviewing the plan and sending you back a blunt, written second opinion. That is usually enough to decide whether we are a fit for each other.