Google I/O Reality Check

I/O 2024

46 announcements tracked. Reality Score: 78% shipped substantially as promised.

Shipped · 31Shipped late · 2Scaled back · 9Rebranded · 3Vapor · 1

Shipped 31

Android I/O 2024

Android 15

Shipped Shipped 2024-10
Promised

Next major Android release with Private Space, Theft Detection Lock, foldable improvements.

Reality

AOSP release September 3, 2024; Pixel rollout October 15, 2024. Shipped on time.

Android I/O 2024

Circle to Search expansion

Shipped Shipped 2024-05
Promised

Available on 100M+ devices, doubling to 200M by end of 2024.

Reality

Expanded steadily through 2024 to Galaxy A series, Galaxy Tab S9 FE, Motorola Razr, Xiaomi 14T, Tecno, Honor and others. By late 2025 it's effectively the default visual-search gesture across modern Android.

Android I/O 2024

Find My Device network expansion

Shipped Shipped 2024-04
Promised

Crowdsourced Bluetooth tracking network across 1B+ Android devices, including offline finding for Pixel 8/8 Pro.

Reality

Launched in US/Canada April 8, 2024 — actually launched ~5 weeks before I/O 2024. Expanded internationally over the following months. Tile/Chipolo/Pebblebee tracker support added.

Was discussed at I/O but had already begun rolling out.

Workspace I/O 2024

Gemini in Workspace (Gmail Q&A, Docs help)

Shipped Shipped 2024-06
Promised

Gemini 1.5 Pro features inside Gmail, Docs, Drive, Slides, Sheets for paying users.

Reality

Shipped for paid Workspace and Google One AI Premium customers. Later (January 2025) Google removed the upcharge and bundled most Gemini features into standard Workspace plans.

Assistant I/O 2024

Gemini Live (mobile conversational AI)

Shipped Shipped 2024-08
Promised

Mobile-first conversational Gemini experience with 10 natural voices, hinted at via Project Astra demo.

Reality

Launched on Pixel 9 in August 2024 as the new default assistant, gated to Gemini Advanced subscribers initially. Made free with camera/screen-share for all Android/iOS users in April 2025.

Workspace I/O 2024

Gemini Q&A, summarize, and Contextual Smart Reply in Gmail mobile

Shipped Shipped 2024-07
Promised

Three Gemini features for Gmail on Android/iOS: summarize threads, contextual smart reply, and Q&A over inbox.

Reality

Summarize and Contextual Smart Reply shipped to Workspace Labs and then Gemini Advanced/Workspace AI users in 2024. Gmail Q&A over inbox shipped on mobile later in 2024.

Developer I/O 2024

Gemma 2 (open weight model, 9B & 27B)

Shipped Shipped 2024-06
Promised

Open-weight LLM at 27B parameters competitive with much larger models, 'launching in June 2024'.

Reality

Released June 27, 2024 on Kaggle and Hugging Face — exactly when promised.

Search I/O 2024

Helpful Content System merged into core ranking

Shipped Shipped 2024-03
Promised

The standalone helpful content classifier folded into Google's core ranking signals.

Reality

Rolled out as part of the March 2024 core update (before I/O), reiterated at I/O. Many independent publishers reported catastrophic traffic losses lasting into 2025–2026.

Defining cause of the 2024 'small publisher apocalypse' narrative on HN.

NotebookLM Audio Overviews (podcast generation)

Shipped Shipped 2024-09
Promised

Early prototype: turn uploaded sources into a podcast-style conversation between two AI hosts.

Reality

Shipped publicly September 2024, became one of Google's surprise viral hits. Interactive Audio Overviews (December 2024), 50+ languages (April 2025), new formats (September 2025).

Arguably the most-loved I/O 2024 product launch.

Developer I/O 2024

PaliGemma (open vision-language model)

Shipped Shipped 2024-05
Promised

Google's first open vision-language model for visual Q&A and captioning.

Reality

Released May 14, 2024 (day of I/O). PaliGemma 2 with 3B/10B/28B variants followed December 2024; PaliGemma 2 mix in 2025.

Cloud I/O 2024

Trillium (6th-gen TPU)

Shipped Shipped 2024-12
Promised

Sixth-generation TPU with 4.7x peak compute per chip vs TPU v5e and 67% better energy efficiency; on Google Cloud 'late 2024'.

Reality

Hit preview in October 2024 and went GA in December 2024 — bang on the promised timing.

Cloud I/O 2024

Trillium TPUs powering Anthropic & external customers

Shipped Shipped 2024-12
Promised

Trillium pitched as the substrate for next-gen frontier model training, including external customers.

Reality

Anthropic publicly committed to using Trillium TPUs at scale through 2024–2025; Trillium also powered Google's own Gemini 2.x training. External availability shipped with Trillium GA in December 2024.

Veo (text-to-video model)

Shipped Shipped 2024-12
Promised

1080p, minute-plus video generation model — Google's answer to OpenAI's Sora.

Reality

Shipped behind a waitlist in VideoFX in 2024; Veo 2 (4K) launched December 2024; Veo 3, 3.1 and Lite followed through 2025–2026. Available in Vertex AI, AI Studio, VideoFX and YouTube.

Initial Veo 1 launch was gated; broad public access came with Veo 2.

Hardware I/O 2024

Wear OS 5

Shipped Shipped 2024-09
Promised

Next Wear OS version with up to 20% improved battery life.

Reality

Shipped September 25, 2024 with Pixel Watch 3 and Galaxy Watch 7 / Ultra. Older Pixel Watch 1 and 2 got it in November 2024.

Shipped late 2

Other I/O 2024

Gemini-generated movie/TV descriptions in Google TV

Shipped late Shipped 2025-01
Promised

Gemini auto-fills missing show/movie descriptions and translates them, implied 'later this year' (2024).

Reality

Slipped past 2024 — broader Gemini-on-TV experiences previewed at CES 2025 and the core auto-description feature only meaningfully appeared in early 2025.

~8 months after I/O 2024 and past the implied within-year timing.

Lyria (music generation model)

Shipped late Shipped 2026-02
Promised

DeepMind's music model powering Music AI Sandbox.

Reality

Lyria stayed mostly behind-the-scenes through 2024. A consumer-facing Lyria 3 rolled out in the Gemini app in February 2026, followed by Lyria 3 Pro in March 2026 (full 3-minute tracks).

Almost two years before consumers got hands-on access.

Scaled back 9

Search I/O 2024

AI-organized search results pages

Scaled back Shipped 2024-05
Promised

AI-categorized results pages for restaurants, recipes, movies and music — 'starting with dining and recipes in the US this week'.

Reality

Briefly appeared in Search Labs for some US users, then quietly de-emphasized in favour of the AI Overviews + AI Mode direction in 2025. Never became the default Search experience as the keynote framing implied.

Eclipsed by AI Overviews fallout and the later AI Mode pivot.

Photos I/O 2024

Ask Photos (Gemini-powered Google Photos search)

Scaled back Shipped 2024-09
Promised

Gemini in Google Photos for natural-language search of your library, launching 'this summer' (summer 2024).

Reality

Slipped: early-access waitlist opened September 2024, broader rollout pushed to 2025. Google paused the rollout in June 2025 over quality issues, then resumed it later that month. Capability also narrower than the keynote demo suggested.

Missed promised 'summer 2024' GA by ~a year and had to be paused mid-2025 for quality — less than promised on both timing and capability.

Hardware I/O 2024

Gemini on Pixel Buds Pro 2 / wearables

Scaled back Shipped 2024-08
Promised

Hands-free Gemini conversations through earbuds previewed in the Project Astra direction.

Reality

Pixel Buds Pro 2 shipped August 2024 with Gemini integration; conversational/Live support on earbuds rolled out incrementally through 2025 but remained more limited than the fluid Astra demo.

Hardware shipped; the seamless Astra-grade earbud experience did not.

Maps I/O 2024

Geospatial AR content in Google Maps

Scaled back Shipped 2024-07
Promised

AR experiences in Maps, piloting in Singapore and Paris later in 2024.

Reality

Paris pilot launched July 2024 with Ubisoft + Google Arts & Culture. Singapore content followed. Remained a small partner-led pilot — never broadened materially.

Barely talked about after launch.

Music AI Sandbox (with YouTube)

Scaled back Shipped 2024-05
Promised

DeepMind+YouTube tool suite for AI music generation, demoed with Wyclef Jean, Justin Tranter, Marc Rebillet.

Reality

Limited rollout to selected artists only; 'broader access' in 2025 still meant invited musicians. Never opened to the general public despite the broad keynote framing.

Pitched as a broader creative tool but remains artist-invite-only as of 2026.

Assistant I/O 2024

Project Astra (universal multimodal AI assistant)

Scaled back Shipped 2025-04
Promised

A 'universal AI agent' demoed with a phone camera and prototype glasses understanding the world in real time.

Reality

Remained a research prototype. Capabilities (live video, screen share, memory) were folded into Gemini Live starting April 2025. Glasses still in trusted-tester program in 2026; no consumer Astra product.

The viral demo was the splashiest I/O 2024 moment — but the cohesive 'Astra' product never shipped as such.

VideoFX (Veo-powered Labs tool with Storyboard)

Scaled back Shipped 2024-12
Promised

Labs tool to turn text into video clips, with a Storyboard mode for scene-by-scene iteration plus music.

Reality

Initial Storyboard feature was cut before public release; VideoFX shifted to single-video generation only. Wider access came via Veo 2 in December 2024.

The promised Storyboard mode never shipped publicly.

Rebranded 3

LearnLM (education-tuned model family)

Rebranded Shipped 2024-06
Promised

Family of Gemini-based models fine-tuned for education, integrated into Search, YouTube, Classroom.

Reality

Available in AI Studio as an experimental model. At I/O 2025, Google folded LearnLM capabilities directly into Gemini 2.5 — no more standalone LearnLM listing.

Now part of mainline Gemini rather than a separate model.

Search I/O 2024

Multi-step reasoning, meal & trip planning in Search

Rebranded Shipped 2024-05
Promised

Multi-step reasoning, meal plans and trip plans available in Search Labs, with date-night and workout plans coming later in 2024.

Reality

Launched in Search Labs for U.S. English users in May 2024. The standalone planning tools were largely subsumed into the broader AI Mode launched March 2025.

Effectively absorbed into AI Mode in 2025.

Vapor 1

Illuminate (research papers to AI podcasts)

Vapor
Promised

Experimental tool turning research papers into AI-narrated audio conversations.

Reality

Available as a free Labs experiment with limited scope (primarily arXiv CS papers). Effectively cannibalised by NotebookLM's Audio Overviews. Still 'experimental' two years later with no visible roadmap or graduation from Labs.

Two years stuck in Labs with no meaningful progress — fails the 12-month public-progress bar.