What to Build Before WWDC 2026: A Developer's Opportunity Map for iOS 27

Every outlet will tell you what Apple is going to announce at WWDC 2026. This is the other question — the one that actually pays rent: given what's coming, what should you build?

We sifted the full rumor pile from a developer's seat, not a consumer's. Most of it is noise if you ship software. A small slice rewrites the cost structure of an entire category of app. Here's the signal, ranked by whether it changes a build decision — with our confidence level stated plainly, because these are rumors and you're about to make bets on them.

A surprising amount of the WWDC 2026 cycle is consumer polish. Useful to know, irrelevant to your roadmap:

• 1Revamped AirPods settings, Maps over satellite, default non-AirPlay casting (e.g. Google Cast). Nice quality-of-life wins. They open no new app category.
• 2Genmoji / Image Playground quality bumps. Real, but incremental — and creative generation in most production apps already runs server-side, so an on-device polish pass rarely moves the architecture.
• 3*The Siri visual redesign (dark theme, Dynamic Island presence). It's the headline screenshot. For builders it matters only through the plumbing* underneath it — which is rumor #2 below, not the paint.

If your reading of WWDC stops at that list, you learned nothing that should change what you ship. Keep going.

These are the ones to plan around. One lowers a cost wall that currently keeps a whole category of app expensive to build; the other hands you a new distribution surface for free.

1. On-device Foundation Models, going multimodal — the cost wall that might fall

Here's the state of play. Apple shipped the Foundation Models framework at WWDC 2025 (iOS 26): direct Swift access to the ~3B-parameter on-device model behind Apple Intelligence, with guided generation, tool calling, and sessions. Genuinely useful — but text-in, text-out. No image input for third-party developers.

That single limitation is why a small army of apps does something that looks insane from the outside: they bundle a 1–2GB vision-language model (MiniCPM-V, Qwen-VL, LLaVA-class) compiled through llama.cpp, shipping it inside the app binary or downloading it on first launch, just to caption or reason over an image privately, on-device. It's a lot of weight, a brutal install-conversion tax, and a per-chip performance floor — all to route around the fact that Apple's own on-device model won't look at a picture for you.

The iOS 27 rumor that matters: reporting points to multimodal input — voice, text, images — becoming standard in this year's on-device AI stack, alongside a reframing of Core ML toward a broader "Core AI" framework with stronger local-inference hooks. Confidence: medium. Multimodal Siri is well-sourced; a public, third-party image-input API on Foundation Models is the specific thing that is not confirmed — and it's the whole ballgame.

So it cuts both ways. If you've been waiting to build private-vision features because the on-device VLM was too heavy, this is your unlock. If your current edge is that you did the hard VLM-bundling work, a system API commoditizes it — and your moat has to move up the stack to the specialized reasoning, not the raw captioning.

2. Siri "Extensions" + App Intents — free distribution you can wire up today

This one is half-shippable right now, which is exactly why it's underrated. Two layers:

• 1App Intents (the part that already exists, since iOS 16). You formally declare the actions your app can perform. Siri and Apple Intelligence can then trigger them directly — "book a ride to the airport" fires your booking flow without the user ever opening the app. This is real today and most apps haven't bothered. That's the gap.
• 2Siri "Extensions" (the rumored iOS 27 layer). Reporting says Apple will open Siri to third-party generative providers (Gemini, Claude) and let installed apps expose generative capabilities on demand through Siri, Writing Tools, Image Playground, and friends. The new Siri becomes a router; your app becomes one of the things it can route to. Confidence: high that App Intents gets a major Siri push; medium on the exact Extensions shape and terms.

The catch worth knowing before you over-invest: Apple hasn't ruled out charging for deeper Siri integration eventually, and some large app makers are hesitating on exactly that uncertainty. The hedge is obvious — App Intents adoption is free, standard, and useful on its own today. Do that now; treat Extensions as upside, not a dependency.

Cross-referencing both rumors against what they actually unlock, and whether you can move now or should wait for June 8 confirmation:

During the keynote and the State of the Union session that follows, these are the specific confirmations that flip a "maybe" into a "build." Print this. Each line tells you what to do the moment you hear it:

• 1"Foundation Models now accepts image input" (third-party, on-device) → the cost wall fell. Greenlight private-vision features; plan to delete bundled-VLM weight.
• 2A named "Core AI" framework replacing/extending Core ML → expect new local-inference APIs in the docs that day. Read the framework reference before competitors do.
• 3App Intents gets a dedicated Siri/Extensions session → distribution surface is real and prioritized. Move App Intents up your sprint.
• 4*Siri Extensions ships with a published API and pricing/terms* → bring-your-own-model assistants become buildable; check whether integration is free before you commit a roadmap to it.
• 5On-device model parameter/quality jump announced → your existing on-device features get better for free; re-test quality-gated features you previously shelved.
• 6Device/chip floor for the new on-device AI → tells you your addressable install base. A high floor (A17 Pro+) changes the unit economics of anything you build on it.

The developers who win the week after a keynote are the ones who did the un-sexy prep before it. None of this is wasted even if the rumors miss:

• 1Adopt App Intents now. It exists, it's free, it's useful with or without the new Siri. Declare your app's 3–5 core actions. This is the single highest-ROI pre-WWDC move and it ships today.
• 2Isolate your AI layer behind an interface. If you bundle or call a model, put it behind a protocol/adapter so swapping in a system API later is a one-file change, not a rewrite. The teams that can swap fastest capture the binary-size and quality wins first.
• 3Inventory what you bundle. Know your exact app-size and first-launch-download cost from any embedded model. If June 8 lets you drop it, you want the before/after number ready for your release notes and your funnel.
• 4Draft the feature you've been postponing because on-device AI was too heavy. Have the spec ready so that if the cost wall falls, you're implementing on June 9, not scoping.

These are rumors. Some will be wrong. The way to play it isn't to assume any single leak is true — it's to make moves that pay off across both outcomes:

• 1Bet on App Intents unconditionally. It's already real. There is no version of WWDC 2026 where adopting it was a mistake.
• 2Bet on multimodal on-device AI conditionally — build the prototype on your own bundled model now so you ship regardless, and architect for a clean swap if Apple hands you the API for free.
• 3Don't bet your roadmap on Siri Extensions terms. Treat it as upside until you've read the actual pricing on June 8.

The pattern underneath all three: Apple's gaps are where indie opportunity lives, and Apple's keynotes are where those gaps open and close. The on-device vision gap is the one to watch this year. Whether Apple closes it for you or leaves it open for you, knowing which happened on June 8 — and having built for both — is the entire edge.

We'll be tracking which apps actually adopt Siri Extensions once the API ships. If that's the kind of signal you want as it lands, that's what we do here.

When is WWDC 2026 and what will Apple announce?

WWDC 2026 runs June 8–12, with the keynote on Monday, June 8 at 10am Pacific. Apple is expected to unveil iOS 27, iPadOS 27, macOS 27, watchOS 27, tvOS 27, and visionOS 27, with a heavy focus on Apple Intelligence — a redesigned Siri, expanded on-device AI, and new developer frameworks. Major hardware is not strongly rumored for the keynote itself.

What's the most important WWDC 2026 announcement for app developers?

Two things, not the consumer features. First, whether Apple's on-device Foundation Models framework gains image (multimodal) input for third-party apps — that would turn the 1–2GB vision models many apps bundle by hand into a system API. Second, the Siri "Extensions" and App Intents push, which lets Siri trigger your app's actions and route generative tasks to it. App Intents is the part you can build on today.

Can I use Apple's on-device AI model in my own app right now?

Yes — the Foundation Models framework shipped with iOS 26 (WWDC 2025) and gives third-party apps direct Swift access to the on-device model for text generation, summarization, structured output, and tool calling. The limitation as of iOS 26 is that it's text-only; image input for developers is the iOS 27 rumor to watch. It requires Apple-Intelligence-capable devices.

What are App Intents and why should I add them before WWDC?

App Intents is Apple's framework (since iOS 16) for declaring actions your app can perform so Siri and Apple Intelligence can trigger them directly — without the user opening the app. Adopting it is free, standard, and useful today, and it positions your app for the rumored Siri "Extensions" system in iOS 27. It's the highest-ROI thing a developer can do ahead of WWDC 2026 because it pays off regardless of what's announced.

Should I keep bundling my own vision model or wait for Apple?

Build on your bundled model now so you ship regardless, but isolate it behind an adapter so you can swap to a system API cheaply. If WWDC 2026 confirms third-party image input on Foundation Models, you can drop the bundled weight for a smaller binary and free model upgrades. If it doesn't, you've lost nothing. The key is architecting for the swap before you know the outcome.

Will Apple charge developers for Siri integration in iOS 27?

Unknown as of late May 2026. App Intents adoption is free today. For the rumored deeper Siri "Extensions" — including routing to third-party AI providers — Apple has reportedly not ruled out charging eventually, which is why some large developers are hesitating. The safe move is to adopt the free App Intents layer now and wait for the keynote to learn the terms before committing a roadmap to Extensions.