Our Field Trip to Google I/O + A Sit-Down With Sundar Pichai + System Update

Summary of Our Field Trip to Google I/O + A Sit-Down With Sundar Pichai + System Update

by The New York Times

55m•May 22, 2026
Listen to Episode

Overview of Our Field Trip to Google I/O + A Sit-Down With Sundar Pichai + System Update

This episode of Hard Fork is a live-ish recap of Google I/O, a hands-on reaction to Google’s newest AI products, and a wide-ranging interview with CEO Sundar Pichai about Google’s place in the AI race. The hosts also break down several big tech stories in their “System Update,” including the end of the OpenAI/Musk trial, Meta’s restructuring, and a fresh wave of AI-related controversy in publishing.

Google I/O: What Google Announced

The big themes

Google’s keynote was framed around:

  • Agentic AI: tools that can do tasks on your behalf
  • Search redesigns: the company’s biggest changes to Search in years
  • Speed and cost: Google is betting on cheaper, faster models at massive scale
  • Multimodal models: systems that can handle text, images, and video

Standout products and ideas

  • A revamped search experience that Google claims is the biggest change in 25 years.
  • Agentic search / alerts-like automation, where users can ask Google to monitor things like home listings, sports scores, or other updates.
  • Gemini 3.5 Flash, which Google pitched as:
    • much faster
    • cheaper than leading frontier models
    • aimed at high-volume, practical use cases
  • Antigravity, Google’s coding assistant / agentic workflow tool.
  • Spark, an agent meant for everyday users, including scheduling and productivity tasks.
  • A new Omni-style multimodal model that can ingest video, images, and text.

Hosts’ reaction

Their overall verdict on I/O was basically:

  • solid, not revolutionary
  • lots of promises that won’t be meaningful until they’re widely shipped
  • Google seems confident it can win by being fast, cheap, and broadly distributed
  • the event had a surprisingly positive vibe, with little of the AI backlash that might be expected at a big tech showcase

One memorable takeaway from the keynote: Google’s leadership sounded increasingly comfortable speaking in almost apocalyptic terms, including Demis Hassabis’s “foothills of the singularity” line.

First Impressions of Gemini 3.5 Flash

The hosts spent time testing Google’s new model and came away with mixed feelings:

What they liked

  • Very fast
  • Doesn’t hit token limits quickly
  • Useful for high-volume workloads

What disappointed them

  • It did not feel like a major leap in capability
  • Some early users on social media were reportedly disappointed
  • Its pricing, while lower than some frontier models, was also more expensive than the previous Gemini Flash
  • The hosts suspect a more powerful model may exist internally but wasn’t ready in time for I/O

Bottom line

This seems like a model aimed more at scale and efficiency than at wow-factor breakthroughs.

Interview With Sundar Pichai

The interview is the episode’s core substantive section. Pichai gives a candid view of Google’s AI position, including where he thinks the company is strong and where it still has work to do.

Google’s standing in the AI race

Pichai says Google is:

  • at the frontier in some areas
  • behind in others, especially:
    • agentic coding
    • tool use
    • long-horizon tasks
    • instruction following in complex workflows

He emphasizes that the AI race is highly dynamic and that even being a few months off in model release cycles can change public perception dramatically.

Why Google is pushing fast, cheap models

Pichai argues that:

  • speed and efficiency matter because Google serves billions of users
  • cost and distribution are strategic advantages
  • Google’s business model supports deploying AI broadly and economically

Search is changing, but slowly

On the future of search, Pichai says:

  • Google wants to bring users along gradually
  • the classic search interface and sources/links are not going away anytime soon
  • the company is evolving toward AI mode and agentic search, but methodically

AI, jobs, and productivity

Pichai’s core argument is that AI will:

  • change the starting point for many jobs
  • make people more productive
  • free up time for higher-value human work

He uses examples like:

  • spreadsheets transforming finance
  • AI helping doctors spend more time with patients
  • AI increasing the amount and complexity of medical and radiology work

AI backlash and public anxiety

Pichai acknowledges that:

  • people are understandably anxious
  • the pace of change is unusual and hard for society to absorb
  • the industry needs to do more to show tangible benefits

He sees the backlash as a natural reaction to a profound technology shift rather than a sign that progress should slow.

Regulation and government oversight

On AI regulation, Pichai seems open to:

  • cross-government coordination
  • security-focused oversight
  • an approach that balances innovation and guardrails

He does not fully endorse heavy pre-release approval systems, but he’s broadly supportive of careful coordination around security and safety issues.

Recursive self-improvement / RSI

Pichai says:

  • models are getting better at coding and orchestration
  • AI can already do impressive things in software development
  • but true recursive self-improvement is not here yet

He argues that if that moment arrives, it should trigger a broader societal conversation, not just an internal lab decision.

Compute and TPUs

Pichai defends Google continuing to sell TPUs and support outside customers:

  • cloud revenue and first-party AI are complementary
  • scale and ecosystem benefits help Google improve its own hardware
  • Google also uses NVIDIA chips, so it is not treating TPUs as an exclusive internal resource

AGI and “the singularity”

Pichai says he believes progress toward AGI is inevitable, but:

  • he avoids overusing the term publicly
  • he thinks the timeline is still uncertain
  • he sees Demis Hassabis’s “singularity” language as essentially shorthand for AGI

System Update

Elon Musk vs. OpenAI: Trial Ends Early

The jury rejected Musk’s claims against OpenAI, but on statute-of-limitations grounds, not on the merits.

What matters

  • The case was dismissed because Musk filed too late.
  • Still, the trial exposed a lot of revealing internal history and generated tons of tech-industry gossip.
  • Musk said he will appeal.
  • The hosts note the irony that Musk himself had previously explored turning OpenAI into a for-profit company and even folding it into Tesla.

Meta’s AI Reorg Continues

Meta laid off 10% of its workforce and reassigned 7,000 employees to AI-related projects.

Takeaways

  • Meta is aggressively reorganizing around AI.
  • The company appears willing to reshape its workforce to catch up in the race.
  • Employee morale seems likely to stay strained.
  • The hosts frame Meta as a company whose core product has become reorganizations in service of AGI ambitions.

The Pope and AI

Pope Leo XIV is expected to release his first encyclical on AI, and Anthropic co-founder Chris Olah is reportedly involved in the launch.

Why it stood out

  • It signals the Vatican’s increasing interest in AI ethics.
  • The hosts joke that this makes the Pope a likely future guest on Hard Fork.

AI Scandals in Publishing

Two literary controversies emerged:

1) A prize-winning short story under AI suspicion

A regional winner of the Commonwealth Foundation’s short story prize was accused of being AI-generated after readers noticed stylistic telltales and AI detector output.

2) A nonfiction book with fabricated AI quotes

A book about truth in the age of AI was found to contain misattributed or invented quotes, apparently generated during an AI-assisted research and drafting process.

The broader point

The hosts argue that:

  • banning AI outright often just encourages people to hide its use
  • AI is increasingly being used for research, notes, and drafting
  • but users still need rigorous human fact-checking, especially in publishing

Hard Fork Live Update

They also announced the lineup for Hard Fork Live in San Francisco, including:

  • Satya Nadella
  • Dylan Field
  • Cindy Cohn
  • Dwarkesh Patel
  • Sayash Kapoor
  • Daniel Kokotajlo

Tickets are sold out, but the conversations will be shared on the podcast.

Main Takeaways

  • Google is leaning hard into agentic AI, Search transformation, and cheap, fast models.
  • The hosts see Google as competent and improving, but not clearly ahead on the highest-end coding and agentic workloads.
  • Sundar Pichai’s position is essentially: Google is strong, the field is moving fast, and the real challenge is turning AI progress into durable, broadly useful products.
  • The broader tech world remains in upheaval, with AI now reshaping law, labor, publishing, and public institutions.