峰会演讲

37:06

The most useful moment is an aside rather than a demonstration: why do all that manual work ourselves, asked immediately after an application crashes. What follows is an argument about which parts of operational work were only ever done by people for lack of an alternative. Troubleshooting is one of the few engineering activities that is genuinely unschedulable — it interrupts whatever was planned and costs attention at the worst moment — so reducing the interruption is worth more than the hours saved. The deployment description contains the detail separating a demonstration from something operable: monitoring with automatic rollback, without which an agent that deploys is a mechanism for producing outages faster. What the session does not address is the failure mode this architecture creates, where a fix addresses a symptom, deploys cleanly and passes monitoring while the real defect remains.

Google Cloud agent platform team / Google I/O

18:05

The distinction Reese draws early is the one that matters: an embodied reasoning model is not a vision model bolted to a robot but the logic unit of the system, fine-tuned on robotics data for spatial understanding, and reasoning about a scene's physics rather than its pixels. That collapses the seam between perception and planning where most traditional robotics failures lived, because the planner no longer receives categories with everything uncategorisable discarded. The browser-based demonstration carries an argument about access as much as capability, since robotics has been gated on hardware and a physics engine in a browser moves the constraint from equipment to ideas. The session's sharpest moment is its last: a model that hallucinates in software produces a strange recipe, and the same error rate attached to something exerting force is a different category of event — a gap of orders of magnitude, not an increment.

Paul Reese / Google I/O

31:34

Pichai's framing of Google as the buffer between people and the raw internet is offered as continuity — search did it, browsers did it, agents do it more — and it is also the most contested claim in the industry, because a buffer decides what passes through. He reaches immediately for the counterweight, the connection people feel to creators they follow, which is precisely the tension the company is currently managing without resolving. Two answers are sharper than the format usually produces. On competition he describes participants running on different pre-training and release cadences rather than at different speeds in one race, which is a more honest account than the leaderboard framing and comes from someone with an interest in leaderboards. On security he acknowledges models improving at cyber work, which is the one domain where better capability does not obviously net out positive, since an attacker needs one vulnerability and a defender needs all of them.

Sundar Pichai / Google I/O

37:59

The anecdote that opens the panel does the work: a robot asked to pick up the extinct animal selected a dinosaur toy, with nothing in its training data connecting the phrase to the object. That transfer from language models into machines with hands is the premise of the current wave. What the practitioners then describe is where it stops. Physical intelligence is about exerting force and using a body to do it, which is knowledge about consequences — the one thing a corpus of internet images contains almost nothing about. The humanoid question gets an honest treatment: not that human shape is optimal, but that the world is already built for it, plus a development-loop argument about collecting data and deploying on the same hardware. The most useful passage is scepticism about the field's favourite shortcut: generated video looks realistic and does not hold up for dexterous manipulation, because looking right and being physically consistent are different properties.

Kenjiro / Google I/O

47:14

The observation that organises this session is not about capability but about attention. Engineers increasingly file a ticket rather than opening an editor, and the code comes back — which changes what the surrounding tools are for. If the agent works while you do something else, the conversation between you is no longer a workspace; it is the mechanism by which the agent surfaces a question it cannot resolve alone. Interfaces built for continuous conversation optimise for flow, and interfaces built for interruption should optimise for the opposite. A runtime constraint follows immediately: an agent that starts a long-running job cannot block until it finishes, which turns out to be a workflow-engine problem rather than a model one. The panel's closing formulation — that deciding what to build is the hard skill and always was — reads as reassurance and functions as a warning, since that judgement is downstream of exactly the work now being delegated.

Tulsi Doshi / Google I/O

10:48

The most honest moment here is an aside about how the presenters have tracked their own projects: plans in documents, plans in spreadsheets, plans in bug comments, and once a plan written on a receipt. That describes the actual category being addressed — not software nobody has built, but the small internal tool every team improvises badly because building it properly was never worth the effort. The demo turns on a single question: the generated app is strictly single player, so what happens when you want to share it with the team? That boundary is where improvised tools historically died, because it is where accounts, shared storage and access rules begin. Here it is crossed in one step, with the access rules generated and deployed automatically — which is convenient, and is also the moment the application acquires obligations nobody reviewed.

Kara Yu / Google I/O

42:48

Four months after Davos, Hassabis put a sharper number on the same forecast: AGI around 2030, give or take a year, arriving gradually rather than as a single moment. His test for it is concrete — a model with a 1901 knowledge cutoff that could produce Einstein's 1905 insights — and by that standard current systems plainly fail. The interview is more useful than the Davos panel on two fronts. First, competitive position: he argues Google's advantage is being the only organisation holding the full stack from chips to billion-user products, citing 900 million monthly users on the Gemini app. Second, method: the AlphaFold story of choosing to fold every known protein at once rather than run a request service is his working example of what acceleration should look like. He closes on a warning aimed at the Bay Area — that direction matters more than velocity, and that the current frenetic pace is not conducive to the deep work the next advances require.

Demis Hassabis / Google I/O

38:13

The clearest statement at I/O of how an agent API differs from a chat API, and the reasoning behind each departure is stated rather than assumed. Three changes matter. Conversation state moves to the server: a call returns an identifier, and passing it back continues the thread, retiring the client-side history array. The data model abandons alternating user and model turns for discrete steps, on the argument that a trace containing reasoning, tool calls, environment responses and compaction was never really a conversation and modelling it as one distorted it. And agents receive their own persistent remote environment rather than acting on the caller's machine — addressable by identifier, and shareable, so a research agent's output files become an application builder's input without passing through the context window. Schmid is explicit that scaffolded environment files are deliberately not model input, which is what keeps large artefacts out of the context budget. Schaeff's first half covers the real-time voice path, where the notable property is speech-to-speech across ninety languages with transcription of both directions.

Philipp Schmid / Google I/O

32:10

The most consequential sentence here concerns what happens after the demo: sometimes you want a quickly assembled application as a proof of concept, and other times you want it to stick around with real users and persistent data. The security additions are best read as answers to the failure modes that creates — keeping prompts off the client, restricting model access to the application, rate limits preventing a few users consuming everyone's quota, ignoring prompts arriving from the client. Each assumption being corrected is one a rapidly built application makes by default, harmless in a demonstration and consequential the first day it is public. The claim that strong security rules were written by the agent deserves the most scrutiny, because an overly permissive access rule produces no visible symptom — it works fine while allowing access it should not.

Firebase Team / Google I/O

45:41

The most instructive fix in this session is not a model change: a response was taking too long and containing too much, so the presenter constrained the prompt to a single sentence under fifteen words and got a very fast response. On a phone, output length is the cost, because every token is generated on the device's own processor while the user waits. That reorganises how these features are built, since the constrained version satisfies the product requirement and the performance requirement with the same change. The first section runs in airplane mode, which settles what is actually being demonstrated. The privacy argument is made about receipts containing card numbers, and the demonstration then reaches for the larger cloud model for the harder version of that task — a tension the session does not comment on. The structural improvement, marking a data class as generable, removes an entire category of runtime failure.

Yolanda / Google I/O

11:30

The most revealing moment in this segment is not the hardware but a question the assistant asks before acting: would you like to confirm. On any other device that is unremarkable; on glasses it states the entire design problem. Every consumer transaction interface built in two decades has relied on a visual review step where errors get caught, and a device worn on the face has no review screen in any meaningful sense. What replaces it is the assistant restating what it is about to do, which must contain everything a user would have checked and be short enough to listen to — two requirements that pull against each other. Adding a tip after confirmation shows this is a conversation with amendments rather than a single transaction, which makes the boundary between negotiating and committing a design decision rather than an emergent property.

Android XR Team / Google I/O

40:18

The transferable content here is not the demonstrations but a set of rules for writing instructions an agent will follow. Supporting files are provided alongside a main instruction file, with explicit statements about when and how each applies, so that material loads at the start but is used only for specific tasks — and so the agent can skip a step entirely when it is unnecessary. One subtask was extracted into its own file specifically so it could be skipped when the user has already named the target, which is instruction design as software design. The practical warning is that customisations should be extracted and renamed, because a bulk update of the shared instructions can silently overwrite them. Two unremarked demo moments establish the real working rhythm: start the long-running task first and fill the time with something else.

Sandhya / Google I/O

40:20

The claim most likely to matter here is about data rather than computation. The measurements behind modern structural biology began accumulating in the 1970s and took roughly fifty years of painstaking work to become the database that made the protein-structure breakthrough possible — and a quantum computer could produce valuable training sets where collecting them experimentally is impractical. That inverts the usual framing, because a dataset does not need a fully error-corrected machine: it needs to be produced once, correctly, and then has permanent value. The technical status report is specific, with coherence times improved roughly tenfold and the remaining obstacles described as system-level engineering rather than physics. The most actionable statement concerns cryptography, where an algorithmic result rather than hardware progress moved the timeline inward.

James Manyika / Google I/O

17:03

The example that justifies this session is a farming application in rural Maharashtra translating agricultural insights and weather warnings into a regional Marathi dialect, offline. That is not a latency argument — the user is unreachable by a hosted model twice over, once through connectivity and once through dialect coverage, and neither problem is solved by a more capable model in a data centre. The most concrete engineering claim concerns speech, where a conventional approach needs a separate 100-to-200-megabyte model per language while the on-device approach needs one small adapter across all supported languages, turning a linear install cost into a nearly fixed one. The guidance to always use the smallest model that reliably solves the use case is the opposite instinct to server-side development, where there is always a later.

Karen / Google I/O

64:55

The chart defining this industry's problem is described in one line: traffic keeps rising while revenue per user steadily creeps down, which the speaker rightly calls the narrative for the sector. The adoption figure offered — a tenfold rise this year with a projection past half of users by 2030 — is real growth from a small base attached to a projection that assumes the curve continues through the part where adoption gets hard, which is where forecasts in this industry have historically failed. The observation about spectrum holdings shifting is easy to skip and shapes more than it appears, because spectrum is the one input that cannot be manufactured or competed into existence, which is why competitive analysis here differs from other technology sectors. The most useful moment is an aside about getting caught up in evolving networks, and the numbers above are why that instinct is correct.

Peter Jarich / MWC Barcelona

151:38

The number establishing how fast this changed is that around seventy per cent of the global telecommunications market now holds at least one non-terrestrial partnership, against a situation twelve months earlier where most operators with anything had exactly one. The framing offered is pragmatic partnership rather than competition, and the word choice does real work: satellite connectivity spent most of its history as a competing answer to the question terrestrial networks answered, and lost comprehensively on cost wherever people actually live. What changed is the question. The coverage gaps identified recur weekly or monthly and are heterogeneous — maritime, remote industrial, disaster response, rural — each with different latency tolerance and willingness to pay, which is why no single answer covers them and why the connected-device case is stronger than the consumer one.

Tim Hatt / MWC Barcelona

58:54

The framing figure should appear in more discussions of AI adoption: in the markets under consideration fewer than twenty per cent of people are using AI at all, with benefit concentrated in developed markets. Every debate about economic impact assumes access, and this session concerns the four-fifths for whom the question has not arisen. Digital literacy is named as the representative constraint, and it matters more than infrastructure because the sequence has already run once — mobile reached these markets faster than forecast, and those who missed out were not the uncovered but those who could not use what coverage delivered. The technical description is a distribution argument rather than a product one, and depends on the layers above being built by people with local knowledge, which the same conditions suppress. The operational discipline is what makes it credible: segmented retention reviewed every week or two, because aggregates rise while intended beneficiaries leave.

GSMA AI for Impact / MWC Barcelona

47:25

The most useful sentence is a diagnosis of how clients are getting it wrong: they see a cost opportunity rather than a revenue one. That framing decides everything downstream, because an organisation treating AI as cost reduction measures success by what disappears and arrives at a smaller version of what it already was. The market number offered — seventeen per cent index growth in 2025 that nobody foresaw — is deployed against bubble anxiety and argues in both directions, since unforecast gains are poor evidence for confidence in any current consensus. The panel then locates the real source of instability outside technology entirely, in three geopolitical situations generating enough uncertainty to dominate planning, which is a corrective worth taking seriously at a technology conference. Their closing question about thriving across multiple futures is correct and sits uneasily beside the cost framing they diagnosed at the start.

Matthew Prince / World Economic Forum Annual Meeting

48:04

The distinction in this session that deserves to travel is a two-word change: the model is moving from an attention economy to an attachment economy. That changes what is measured and what regulation would have to address, because attention competes for time while attachment competes for relationship, and the two produce different products from identical technology. Attention is finite in a way people notice; attachment produces reliance that feels like preference, which makes it harder to regulate for the same reason it is harder to notice. The supporting argument is about incentives rather than intent: earlier engagement produces more data and longer relationships. The regulatory proposal — measuring well-being outcomes rather than asking for safety by design — identifies the right target without solving the measurement problem that made regulators settle for a floor in the first place.

Helen Clark / World Economic Forum Annual Meeting

35:55

Musk names his constraint without hedging: AI deployment is limited by electrical power, with chip production rising exponentially against electricity growing at three to four per cent a year, and a crossover he expects within the year where more chips are made than can be switched on. He then names the exception — China, building nuclear at scale and deploying solar at an order of magnitude beyond everyone else — and the room moves on, though placed against his own framing it is the most consequential thing said. The remainder describes a world where the constraint is solved: robots building robots until human wants saturate, humanoid units on sale to the public within roughly two years, and systems exceeding collective human capability around 2030. One claim deserved scrutiny it did not receive — that orbital compute becomes cheapest within three years — because it contradicts his own position that inference must sit near users.

Elon Musk / World Economic Forum Annual Meeting