About This Week’s Covers

This week’s cover is inspired by my great friend, Phillip, who witnessed quite a spectacle while he was out on a jog. Phillip is an avid birdwatcher, and one of his pet peeves is the fact that house cats absolutely decimate songbirds.His Facebook status tells you what happened:

“Top 5 nature moment as I was running this afternoon when a Red-tailed Hawk (not Red-shouldered) flew just above me with a pet cat in its talons. It’s nature’s food chain seeking it’s justice! If you are going to let Fluffy wander outdoors where he will participate in ecological harm to the tune of killing 2 billion songbirds a year, then you better be prepared to lose him to a gruesome end without saying goodbye. It’s just how it works. Unless you live on a farm, keep your cats indoors.”

I decided to use this as my cover image, and I tried a bunch of variations, from religious stylings to Gonzo illustrations to Robert Crumb to Francis Bacon. Phillip is a Lutheran minister in Richmond, Virginia.

GPT made the image, I face swapped Phillip a second time with InsightFace and layered it with Photoshop. The title logo is handmade to look like the ELCA logo, using Proxima Nova font in Photoshop.

Some of the outtakes were strong!

GPT Image Outtakes

Gemini Outtakes

Here are some of my favorite category covers of the week.

Humanities Reading for The Week

I decided to go with three different humanities readings this week. As usual, I had Claude pick the readings, even though I’m an English major. I usually learn something new (it picks about a dozen, and I choose my favorite).

The first is Ecclesiastes 3:19, since Phillip is a minister. The second is “The Windhover” by Gerard Manley Hopkins. The third is a poem called “Design” by Robert Frost. Each reading ties in to Phillip’s incredible experience while running.

Ecclesiastes 3:19
“For the fate of the sons of mankind and the fate of animals is the same. As one dies, so dies the other; indeed, they all have the same breath, and there is no advantage for mankind over animals, for all is futility.”

The Windhover
Gerard Manley Hopkins
1844 –1889
I caught this morning morning’s minion, king-
dom of daylight’s dauphin, dapple-dawn-drawn Falcon, in his riding
Of the rolling level underneath him steady air, and striding
High there, how he rung upon the rein of a wimpling wing
In his ecstasy! then off, off forth on swing,
As a skate’s heel sweeps smooth on a bow-bend: the hurl and gliding
Rebuffed the big wind. My heart in hiding
Stirred for a bird,—the achieve of; the mastery of the thing!

Brute beauty and valour and act, oh, air, pride, plume, here
Buckle! AND the fire that breaks from thee then, a billion
Times told lovelier, more dangerous, O my chevalier!

No wonder of it: shéer plód makes plough down sillion
Shine, and blue-bleak embers, ah my dear,
Fall, gall themselves, and gash gold-vermillion.

Design
Robert Frost
1874 –1963
I found a dimpled spider, fat and white,
On a white heal-all, holding up a moth
Like a white piece of rigid satin cloth–
Assorted characters of death and blight
Mixed ready to begin the morning right,
Like the ingredients of a witches’ broth–
A snow-drop spider, a flower like a froth,
And dead wings carried like a paper kite.

What had that flower to do with being white,
The wayside blue and innocent heal-all?
What brought the kindred spider to that height,
Then steered the white moth thither in the night?
What but design of darkness to appall?–
If design govern in a thing so small.

This Week By The Numbers

Total Organized Headlines: 676

This Week’s Executive Summaries

For the week ending May 22, I organized 676 links into about 60 categories. 150 links went into the top stories and executive summaries.

I’m seven weeks behind because I took seven weeks off to be with my family!

Everyone is home for the summer, and nothing can replace these years while we still get to have the kids at home. The national dance competition is over now. Both kids are into their summer jobs. I’m going to try to catch up a little bit, and then I’ll probably get behind again in late July and August when we take our oldest daughter back to Colorado.

I organized things this week based on what I wanted to know when I go back and read it. There are some announcements coming up that will have product releases in the coming weeks, so I didn’t put them at the top. Pre-releases and betas are great, but it’s better to wait until they’re fully available to the public.

For now, I’m just going to summarize the big items in the order that I want to remember them. There are 43 top stories in all, below.

Starting with Anthropic news, Andrej Karpathy announced that he’s going back to Anthropic to join some of his old friends from the early days of AI leadership.

Andrej is my favorite thought leader in all of artificial intelligence. Second would be Dr. Jim Fan from Nvidia.

Andrej is going to be working on massive training runs that provide Claude its core knowledge and capabilities. He’s going to build a new team that will accelerate pre-training research and development.

Some benchmarks came out about Claude Mythos that show it absolutely dominates. On the leaderboard scores, the best frontier model was ChatGPT 5.5 at 41%, and Claude Mythos comes out at 69%. Claude Opus 4.7 is at 27%.

So basically, GPT-5.5 and Mythos obliterate the third-place contender, Opus 4.7. The idea that Opus 4.7 gets 27% and Claude Mythos gets 69% is jarring.

Anthropic’s valuation has shot through the roof. In February, the valuation was $350 billion. In May, it was $900 billion.

Anthropic’s annual revenue run rate at the end of the year was $9 billion. By May, five months later, it was at $45 billion. So the valuation was basically 3x in three months, and the annual revenue run rate was 5x in five months.

It’s safe to say that despite pointed attacks from the US government, Anthropic has been pushing back successfully on all fronts: throwing its weight around with Mythos, demonstrating revenue and valuations, and stating that thanks to Mythos, the United States is now 12 to 24 months ahead of China.

Just a couple months ago, all the open-source models were around five to eight months behind and closing in. Anthropic just blew that away, saying we’re one to two years ahead. The government is now working to help stop other countries from distilling US frontier technology.

This has to be the absolute craziest five months out of the past three years.

Switching over to Google, Google announced that they are going to completely rework Search to the point where the search results can actually build multiple agents, create alerts, and write code.

Google also released a multimodal video editing tool called Omni that can handle any kind of input or output. You can edit your videos with it, give it any combination of text, images, or videos, and it can combine or mash up anything.

There is no moat. Especially if you’re building derivative works that are all leveraging your own creations and mashing them together, that’s basically editing or green screening. It’s no longer just generative slop.

You could give Omni assets, logos, and text…and just like you would build a strong image output via Nano Banana, you can do the same with video.

A lot of Google’s releases this week were tied to its big event, Google I/O. One of the announcements was that last year at the event, Google was processing 480 trillion tokens a month. Now they’re processing 3.2 quadrillion tokens a month. That’s seven times more than last year.

In addition, Gemini’s user base has doubled and now surpasses 900 million users.

Robotics company Figure started a live-stream demonstration of its robot sorting packages on an assembly line, hoping to hit eight hours of autonomous work. Then it crossed the 30-hour mark. Then it was six days later. And then seven days. By the end of this week, it had been running for over 144 hours straight, and it had sorted 180,000 packages. Then it entered the eighth day without a single error. As of my cutoff for this newsletter, it was still going!

Anthropic

Karpathy joins Anthropic to lead autoresearch pre-training team

Personal update: I’ve joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
https://x.com/karpathy/status/2056753169888334312

Karpathy was literally hired for RSI he’s starting a new pre-training team at Anthropic that focuses on autoresearch per Axios
https://x.com/scaling01/status/2056773883982762114

Claude Mythos dominates agent-control benchmark ahead of GPT-5.5

It’s over Mythos is insane
https://x.com/scaling01/status/2055362921803211248

Anthropic valuation nears $900 billion as revenue quintuples in five months

Anthropic in a nutshell: Valuation: February 2026: $350B May 2026: $900B Almost 3x in just 3 months. ARR: End of 2025: $9B End of May 2026: $45B Almost 5x in 5 months. Holy. via FT
https://x.com/kimmonismus/status/2055222524774846576

Anthropic: U.S. AI lead over China may have jumped from 8 months to 12–24 months

Not only is Anthropic saying they will have a “”country of geniuses in a datacenter”” by 2028, but also that the US could be ahead by 12-24 months. Before GPT-5.5 and Claude Mythos chinese labs were ~8 months behind in broader capabilities and ~5 months behind in coding. However,
https://x.com/scaling01/status/2055018330365345896

Google

Google turns Search into a free agentic coding platform

Agentic coding on the scale of search – this summer for everyone, free of charge #google That’s insane. Google Search will be able to help you build. I mean – I don’t think we yet are realizing what it means on that scale.
https://x.com/TheTuringPost/status/2056795871098913209

How information agents work in Search: 1️⃣ Start with a total brain dump of what you want your agent to keep you updated on 2️⃣ The agent will break down your complex question and map out a plan 3️⃣ It will determine the urgency — understanding that you need in-the-moment intel
https://x.com/Google/status/2056794675214700764

Introducing our brand new, intelligent Search box — totally reimagined with AI. This is the biggest upgrade to our Search box in 25 years and it’s starting to roll out today. Designed to anticipate your intent, the new Search box helps you formulate your question with AI-powered
https://x.com/Google/status/2056793802141044786

Soon, you’ll be able to create and manage multiple AI agents for your many tasks — right in Search ✨ We’re starting with information agents: 🔹These agents intelligently look across everything on the web, including blogs, news sites and social posts, plus real-time data on
https://x.com/Google/status/2056794282502054066

This is mind blowing actually. I’m not that easy to impress but this will change everything. This is the second time Google is changing how people search information and what they get as a result. To clarify: Google is integrating the agentic coding capabilities of Gemini 3.5
https://x.com/Kseniase_/status/2056798225378783656

Today we are starting to roll out the biggest upgrade to the Google Search box in over 25 years — now completely reimagined with AI, along with Gemini 3.5 Flash as the new default model for AI mode users globally!
https://x.com/OfficialLoganK/status/2056802276124328352

Google launches Gemini Omni, a multimodal video generation and editing model

Gemini Omni is a major leap in world understanding & multimodal editing! It can take photos, video & audio and build entirely new scenes. Over time it’ll be able to handle any input & any output – starting w/ video You can even give it your own videos & iterate on your ideas:
https://x.com/demishassabis/status/2056831486251380783

Gemini Omni is coming to the Gemini app for paid subscribers today. It lets you bring your ideas to life using any combination of text, images, and video inputs. Just open up Gemini, attach a video from your camera roll, and change it around. It’s that simple. #GoogleIO
https://x.com/GeminiApp/status/2056800579159216202

Gemini Omni is our new AI model that can create anything from any input, starting with video. 🪄 Hear from @DemisHassabis on how you can mix text, audio, and images to generate and edit high-quality videos just by having a conversation. #GoogleIO
https://x.com/Google/status/2057180052979409172

Gemini Omni is so fun – insanely great at editing videos!
https://x.com/joshwoodward/status/2056827449556845051

Gemini Omni is wild
https://x.com/osanseviero/status/2056863263305105424

Gemini Omni was bigger news than Gemini 3.5 Flash
https://x.com/scaling01/status/2057143531622334678

Gemini Omni: “”a dramatic reading of Death by Water from the Wasteland by a man eating garlic bread while balanced on a unicycle on a small platform over a churning sea of tomato sauce in which, at the center, sites a meatball with bright blue eyes wearing a top hat””
https://x.com/emollick/status/2056791733619376315

I had early Gemini Omni access: “”sea otter in a pilot’s uniform explains why Spirit Airlines went bankrupt to a river otter who is distracted by their laptop while they are in a hot air balloon over NYC. in the next balloon over, william shakespeare fights a robot made of pizza””
https://x.com/emollick/status/2056788122369712148

Introducing Gemini Omni 🔮…….. Omni is our new model that can create anything from any input — starting with video (think Nano Banana but for video). Available in the Gemini App, Flow, and YouTube, with API support coming soon!
https://x.com/OfficialLoganK/status/2056787874260164628

Meet Gemini Omni — our new AI model that can create anything from any input, starting with video. 🪄 #GoogleIO
https://x.com/Google/status/2056786395067552140

Nano Banana for video is here! Google has long touted that Gemini is natively multi-modal in & out — but Omni is the first glimpse into the power of that paradigm applied to creation. Toss in a video and do multi-turn edits. Toss in audio and get reactive visuals. It’s kinda
https://x.com/bilawalsidhu/status/2056790381514076449

Starting today, Gemini Omni is rolling out to all Google AI Plus, Pro and Ultra subscribers globally at https://t.co/382WL5xSvc and in the app. Soon, we will support more output formats, like image and audio. Give Gemini Omni a try and share your creations in the replies. 👇
https://x.com/GeminiApp/status/2056814117047132301

The Odyssey and the Iliad get so many movie treatments but the sequel, the Roman Aeneid, is entirely ignored. Here is a teaser trailer from one prompt to Gemini Omni. The first pass made all the flags Danish(?) but Omni is capable of editing video, so I asked for their removal.
https://x.com/emollick/status/2056855332127711387

The real „wow“ moment is Gemini Omni. A world model towards AGI. It can create anything from any input. This is insane.
https://x.com/kimmonismus/status/2056802929957568881

Trying Gemini Omni on real life #google
https://x.com/TheTuringPost/status/2057167259877916679

We’re dropping Gemini Omni: our first step towards a model that can create anything from anything – starting with video. It combines Gemini’s intelligence with our generative media systems – representing a leap forward in world understanding, multimodality, and editing 🧵
https://x.com/GoogleDeepMind/status/2056786446636212467

Big improvements in quality and capabilities from the Omni team, excited to see how we translate this to progress for embodied AGI
https://x.com/jparkerholder/status/2056789448554062232

By default Omni has a bit of a professional “”look””, but can also create “”normie”” videos with the right prompting.
https://x.com/shlomifruchter/status/2056858151987884087

Creating, remixing, and editing a video is easier than ever with Gemini Omni. It offers a fluid, conversational way to create and edit. Just upload a video from your camera roll and ask Gemini to make changes.
https://x.com/GeminiApp/status/2057159933934907825

Edit your own videos with Gemini Omni with just a conversation. 🎥 Prompt the changes you want to see to reimagine the action, change the point of view, or adjust the lighting over multiple turns. Every instruction builds on the last, so your characters stay consistent, the
https://x.com/Google/status/2056786888930062369

Gemini Omni can create anything from any input, starting with video. 🪄 This means you can combine images, audio, video and text as input and generate high-quality videos. Or use drawings to create in a way that matches your vision. #GoogleIO
https://x.com/Google/status/2056786781992071172

Gemini Omni combines an intuitive understanding of physics with Gemini’s real-world knowledge and reasoning. 🌐 Now, the stories and outputs you create won’t just look photorealistic — they’ll behave like the real world. #GoogleIO
https://x.com/Google/status/2056786589175677089

Gemini Omni Flash is rolling out starting today. Here’s where you can find it: 🔹 Today: Google AI Plus, Pro and Ultra subscribers globally in the @GeminiApp and @FlowbyGoogle . 🔹Rolling out starting this week, for no cost: @YouTube Shorts and the YouTube Create app.
https://x.com/Google/status/2056789307856462061

Gemini Omni Flash: > a recording from a capsule on the london eye, a jerky zoom into something in the distance and then refocusing (with a bit of back and forth) (no timestamp or dialog) Note the world knowledge of London’s landscape, and the way the video is gently moving like
https://x.com/fofrAI/status/2056789242274259242

Google’s AI usage explodes to 3.2 quadrillion tokens monthly

At last year’s #GoogleIO, we were processing 480 trillion tokens a month across our surfaces. Now, we’re processing over 3.2 *quadrillion* tokens a month. That’s a 7x increase in just a year. These tokens represent problems being solved — by a user or a developer or a
https://x.com/Google/status/2056783102085640252

In just a year, @GeminiApp users have more than doubled, surpassing 900 million. #GoogleIO
https://x.com/Google/status/2056783643381543253

Government

California orders statewide plan for AI workforce disruption

Governor Newsom signs first-of-its-kind executive order to prepare workers and businesses for potential AI disruption | Governor of California
https://www.gov.ca.gov/2026/05/21/governor-newsom-signs-first-of-its-kind-executive-order-to-prepare-workers-and-businesses-for-potential-ai-disruption/

Robotics

Figure AI’s humanoid robots sort 249,560 packages in weeklong livestream

76,940 packages over 61 hours 18 minutes The robots are averaging roughly 1 package every 2.9 seconds No breaks
https://x.com/adcock_brett/status/2055534794667442361

Day 2 is Live: Watch humanoid robots Bob, Frank, and Gary running 24/7. This is fully autonomous running Helix-02
https://x.com/adcock_brett/status/2054970993442169230

Figure is now entering Day 8 of the most boring video we’ve ever posted F.03 has been working fully autonomously, non-stop for a full week – and we’ve livestreamed the entire thing
https://x.com/adcock_brett/status/2057145678355460126

Figure just live-streamed 8 hours of fully autonomous, unsupervised work We just huddled internally – and guess what? We’re not stopping here 24/7 LIVESTREAM
https://x.com/adcock_brett/status/2054729581391962353

Figure just set a new standard in unedited humanoid demos. They completed 24 hours of autonomous work on a live stream, where three Figure 03 robots took turns processing delivery packages. At a rate of ~21 packages/minute, that is less than 3 seconds per package.
https://x.com/TheHumanoidHub/status/2054993551973589304

Figure launched yesterday aiming to hit 8 hours of autonomous work We’ve now crossed 30 hours of continuous operations with no downtime 🔥🔥🔥 We also welcomed Rose to the team – helping push us past 38,000 packages processed
https://x.com/adcock_brett/status/2055075231002407417

Figure livestream is back for day 6! Aime got some well deserved sleep, while F.03 has sorted another 17,444 packages
https://x.com/adcock_brett/status/2056420115701448915

Livestream Day 4 has begun. We’ve been running F.03 humanoid robots 24/7 – fully autonomous, no breaks, and no downtime
https://x.com/adcock_brett/status/2055695727985352722

Silicon Valley’s Latest Binge-Watch Is a Humanoid Warehouse Worker – Business Insider
https://www.businessinsider.com/figure-ai-turned-a-humanoid-sorting-packages-must-see-tv-2026-5

The Internet can’t stop watching Figure AI’s humanoid robots handling packages – Ars Technica
https://arstechnica.com/ai/2026/05/the-internet-cant-stop-watching-figure-ais-humanoid-robots-handling-packages/

This week, Figure started an 8 hour livestream to demonstrate fully autonomous, nonstop work We’ve now crossed 50 hours of continuous operation with no downtime, sorting over 63,000 packages The robots are still running – we’re going until failure 🔥🔥🔥
https://x.com/adcock_brett/status/2055376836218310926

We’re entering Day 7! We’ve been running close to a week 24/7 and fully autonomous We’ve sorted over 180,000 packages and have been running continuously, without failure, for 144 hours
https://x.com/adcock_brett/status/2056786783539692023

Anthropic

Anthropic lets enterprises run Claude agents inside their own perimeter

New in Claude Managed Agents: self-hosted sandboxes and MCP tunnels | Claude
https://claude.com/blog/claude-managed-agents-updates

Anthropic and Gates Foundation launch $200M AI-for-good partnership

Anthropic forms $200 million partnership with the Gates Foundation \ Anthropic
https://www.anthropic.com/news/gates-foundation-partnership

Anthropic commits $45 billion to xAI’s Colossus supercomputers

SpaceX also disclosed exactly how much their deal with Anthropic is worth. They state that they have “”entered Cloud Services Agreements with Anthropic in May 2026 for capacity on COLOSSUS and COLOSSUS II at $1.25B/month through May 2029″” that’s ~$15B/year, and an ~$45B TCV over
https://x.com/SemiAnalysis_/status/2057218890288030110

We’re expanding our partnership with @SpaceX, and will be scaling up on GB200 capacity in Colossus 2 throughout June. Appreciate @elonmusk and the team helping us find good homes for the Claudes.
https://x.com/nottombrown/status/2057194829986300375

Wow, Anthropic has agreed to pay @SpaceX $1.25 billion per month through May 2029 for AI compute capacity, with capacity ramping in May and June 2026 at a reduced fee. Huge deal
https://x.com/SawyerMerritt/status/2057208968741560744?s=20

Ethics

Rising bond yields threaten AI industry’s massive debt-fueled buildout

The 10-year US Treasury just hit 4.568%, highest in 10 months. The 30-year is back above 5%. Inflation is re-accelerating, markets have fully priced out any Fed rate cut this year, and some are now betting on a hike. Meanwhile, the AI industry is running the largest debt
https://x.com/kimmonismus/status/2055293526125232332

Frontier agents are not fully rogue-ready, but METR says they are getting close

The full Frontier Risk Report includes more than we can describe here, with over 200 pages of appendix materials detailing how the pilot worked, what we asked participants, and what we learned, along with transcripts from our evals. See our website:
https://x.com/METR_Evals/status/2056800047258649049

Data centers boost local incomes but drive up housing and power costs

“Data centers create economic activity, especially in directly related sectors and during construction, and they are associated with larger county-level income aggregates. They also raise electricity prices and are associated with higher house prices,”
https://x.com/emollick/status/2056136463201300572

Artist tricks X users into slamming real Monet as AI slop

An artist just put a mirror to a lot of society’s knee-jerk critiques of AI, using Claude (Monet). SHL0MS, known for his viral experiments, posted an image of a real Monet water lilies painting on X and told people it was AI-generated. Hundreds of replies poured in explaining
https://x.com/TheRundownAI/status/2054942919367197112?s=20

Pope Leo XIV to issue landmark encyclical confronting AI

Vatican prepares for Pope Leo XIV’s AI-focused first encyclical | AP News
https://apnews.com/article/vatican-artificial-intelligence-pope-musk-nvidia-trump-889c0066f0d5ce784c07abb72b33e24c

Google

Google launches standalone Antigravity desktop app for AI agents

Say hello to the new standalone desktop application for Google @Antigravity. 💻 It’s agent-first — focusing on core agent conversations, agent-produced artifacts, and multi-agent orchestration. #GoogleIO
https://x.com/Google/status/2056788868092006891

Google’s Gemini launches Daily Brief agent for morning digests

Introducing our newest out-of-the-box agent, Daily Brief. It creates a personalized digest in @GeminiApp that’s designed to be your first stop every morning. It synthesizes information from your inbox, your calendar and your tasks to find the most important things for you to be
https://x.com/Google/status/2056801159071883342

Daily Brief is a new personalized digest that’s designed to be your first stop every morning. It gathers info from your inbox, calendar, and tasks to prioritize, organize, and suggest the next steps for you in a super concise morning digest that’s built for skimming. #GoogleIO
https://x.com/GeminiApp/status/2056800978343764238

Get a head start on your day with Daily Brief. Gemini can now proactively flag what matters most in an easily digestible to-do list, so you’re ready for the day before you even finish breakfast.
https://x.com/GeminiApp/status/2057500470147698936

Google’s Gemini-powered audio smart glasses launch this fall

Intelligent eyewear with Gemini is coming this fall
https://blog.google/products-and-platforms/platforms/android/android-xr-io-2026/

Google launches Gemini 3.5 Flash for agentic workflows

Gemini 3.5: frontier intelligence with action
https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-5/

Gemini 3.5 feels like the start of a new era for Gemini, we spent the last 2.5 years putting the infrastructure, products, team, etc in place (learning lots of lessons along the way). The model is the product, please keep the feedback coming!
https://x.com/OfficialLoganK/status/2057104170310902166

Gemini can now connect to even more apps, including @OpenTable, @Canva, and @Instacart. Whether you’re booking a table at a restaurant, creating a flyer, or ordering groceries, Gemini doesn’t just find info, it helps you take action seamlessly with connected apps.
https://x.com/GeminiApp/status/2057550225863246236

(deliberately not hyping Gemini 3.5 Flash too much this time. looks like an insane model, but you know how it is with self-reported benchmarks)
https://x.com/scaling01/status/2056794370909593987

@demishassabis @antigravity @GeminiApp Are you aware that running AA intelligence index cost almost 2x with 3.5 Flash than it did with 3.1 Pro?
https://x.com/giffmana/status/2057155343390494949

⚡ Gemini 3.5 Flash is now available in @code. Give it a try!
https://x.com/code/status/2056803208559759447

📣 @GoogleAI’s Gemini 3.5 Flash is now generally available and rolling out in GitHub Copilot. Early testing shows ➡️ It has strong tool use, fast response times, and high cache efficiency ➡️ It is it well-suited for fast, iterative agentic coding workflows Try it out in @code.
https://x.com/github/status/2056801675042779279

1/ Today at #GoogleIO, we’re releasing Gemini 3.5, our latest family of models combining frontier intelligence with action. We’re starting by releasing 3.5 Flash, which is built to help you execute complex, long-horizon agentic workflows. Gemini 3.5 Flash is our strongest model
https://x.com/JeffDean/status/2056793419033588091

A closer look at Gemini 3.5 Flash by @GoogleDeepMind In the Code Arena: Frontend we see sweeping gains, and a Flash model now surpasses the previous Pro variant. – vs. 3 Flash, a +70 jump overall, large improvements in every subcategory – vs. 3.1 Pro, outperforms it in every
https://x.com/arena/status/2056803661859479812

Also had some early access to Gemini 3.5 Flash. Very fast for a flash model and very capable, though not as powerful as a full frontier model. I added it to the gallery or procedurally generated one-shot towns (it made one error that it corrected):
https://x.com/emollick/status/2056798490353705380

Artificial Analysis benchmarks were featured in yesterday’s Gemini 3.5 Flash launch Yesterday @GoogleDeepMind released Gemini 3.5 Flash at Google I/O ’26 and our benchmarks were used by @sundarpichai to highlight the model’s leading position on the Intelligence vs. Speed Pareto
https://x.com/ArtificialAnlys/status/2057181290412261557

Clearly these people haven’t experienced the magic of Gemini 3.5 Flash Preview on High in the new Antigravity CLI
https://x.com/theo/status/2056826014739890204

Gemini 3.5 Flash can detect and reason I asked it to detect and label cars on each lane replacing a custom model + post-processing pipeline with a single API call
https://x.com/skalskip92/status/2057502215506473121

Gemini 3.5 Flash delivers fast, consistent performance that rivals other leading models – at a fraction of the price. It can plan and reason across massive codebases and deploy subagents to work in parallel over a long horizon. It outperforms 3.1 Pro on coding and agentic
https://x.com/GoogleDeepMind/status/2056787990110994511

Gemini 3.5 Flash delivers sustained frontier-level performance at lightning-quick speeds: Beats 3.1 Pro on coding & agentic benchmarks Terminal-Bench 2.1 (76.2%), GDPval-AA (1656 Elo) and MCP Atlas (83.6%) 4x faster than other frontier models (12x in Antigravity!) SOTA on
https://x.com/koraykv/status/2056795667088204234

Gemini 3.5 Flash has landed #9 for Text and Code Arena: Frontend. Code Arena: Frontend evaluates models on agentic frontend coding tasks from real users building apps and websites (HTML and React). Scoring 1507, this is a significant +70 point improvement over Gemini-3 Flash.
https://x.com/arena/status/2056793176720195693

Gemini 3.5 Flash is an insane step forward. Google now has to thread the needle and make a great agentic coding product akin to Codex/Claude Code. Does Gemini CLI even exist? Oh it’s now Antigravity CLI? Ugh… Google being Google.
https://x.com/zachtratar/status/2056848643580482002

Gemini 3.5 Flash is available now globally to: ✨ Everyone: @GeminiApp and in AI Mode in Google Search ✨ Developers: In Google @Antigravity and Gemini API in @GoogleAIStudio & @Android Studio ✨ Enterprises: In Gemini Enterprise Agent Platform We’re just getting started —
https://x.com/Google/status/2056791527314387208

Gemini 3.5 Flash is now GA. Our most capable Flash model, built for agentic execution, coding, and long-horizon tasks. – Outperforms Gemini 3.1 Pro on coding and agentic tasks – 1M token context window with 65k max output tokens – 4x faster output tokens/sec – 4 thinking levels:
https://x.com/_philschmid/status/2056794978517750165

Gemini 3.5 Flash is our strongest agentic and coding model yet, outperforming Gemini 3.1 Pro on challenging coding and agentic benchmarks.
https://x.com/Google/status/2056788281317306466

Gemini 3.5 Flash is starting to roll out to everyone globally, free of charge. Try it by selecting “3.5 Flash” from the model dropdown menu at https://t.co/382WL5xSvc or in the app.
https://x.com/GeminiApp/status/2057140474192994356

Gemini 3.5 Flash Pricing confirmed at $1.5 / $9 per mtoks
https://x.com/scaling01/status/2056793465715822720

Gemini 3.5 Flash ranks #1 on the APEX-Agents-AA benchmark, outperforming much larger models a whole size above it.
https://x.com/OfficialLoganK/status/2057460544643404125

Gemini 3.5 Flash scores kinda low on the Coding Index due to terrible TerminalBench-Hard scores
https://x.com/scaling01/status/2056796392899645919

Google optimized Gemini 3.5 Flash to make it run up to 12x faster (~867 tokens/s) than comparable models in AntiGravity
https://x.com/scaling01/status/2056790573961326680

Google’s new Gemini 3.5 Flash is the clear leader on the Intelligence vs Speed Pareto frontier and makes large gains on GDPval-AA (real-world agentic tasks), but is 5x the cost of Gemini 3 Flash @GoogleDeepMind gave us pre-release access to Gemini 3.5 Flash, the latest model in
https://x.com/ArtificialAnlys/status/2056795055512596817

GPT-5.5-medium has lower end-to-end latency, uses less tokens and is overall smarter and cheaper than Gemini 3.5 Flash it might genuinely be over for anyone not named OpenAI or Anthropic
https://x.com/scaling01/status/2056803273756000721

Insane evals for a Flash model! Gemini 3.5 Flash is really good for its size!
https://x.com/kimmonismus/status/2056791681073316071

Introducing Gemini 3.5: our newest family of models combining frontier intelligence with real-world action. The first release is 3.5 Flash, our strongest model yet for agents and coding 🧵
https://x.com/GoogleDeepMind/status/2056787987774816525

Just off stage at #GoogleIO, some highlights from this morning 🧵 Gemini 3.5 Flash is available today for everyone in @antigravity and across our products and APIs. Compared to 3.1 Pro, 3.5 Flash is better across almost all benchmarks with huge progress in coding. It’s also
https://x.com/sundarpichai/status/2056796893951426705

Massive updates to @GoogleAIStudio and the Gemini API 🤯 – Gemini 3.5 Flash! – managed agents so you can easily build agentic products with the antigravity harness – native Android app creation right in AI Studio – native workspace integrations – 1 click export to antigravity
https://x.com/OfficialLoganK/status/2056840620610830638

Meet Gemini 3.5 Flash — our strongest agentic and coding model yet. It delivers frontier-level performance at 4x the speed of comparable frontier models — often at less than half the cost. Generally available, starting today. 🧵 #GoogleIO
https://x.com/Google/status/2056788266872140232

My notes on Gemini 3.5 Flash – 3x the price of Gemini 3 Flash but Google are planning to use it for many of their own products
https://x.com/simonw/status/2056867815605625172

Our new Gemini 3.5 Flash is our strongest agentic and coding model yet, delivering frontier-level performance at 4x the speed of comparable frontier models at less than half the cost. We had @TulseeDoshi break down what you need to know about it.
https://x.com/Google/status/2057257773868388448

some more Gemini 3.5 Flash benchmarks by Artificial Analysis AI: – the APEX-Agents-AA score is excellent – expected higher on CritPt – reasoning efficiency could also be better, but it kind of depends on what setting they used. if it’s the max then it’s very good – Price/Perf
https://x.com/scaling01/status/2056798645983334890

Starting today, Gemini 3.5 Flash is rolling out to everyone globally, free of charge. Try it by selecting “3.5 Flash” from the model dropdown menu at https://t.co/382WL5xSvc or in the app, and let us know how you’re using it in the replies.
https://x.com/GeminiApp/status/2056789742910595342

TPUs are insane Gemini 3.5 Flash is running at ~867 tokens/s almost as fast as Kimi-K2.6 on Cerebras custom chips
https://x.com/scaling01/status/2056791726677782743

We asked our agents to build a working operating system from scratch using @Antigravity 2.0 and Gemini 3.5 Flash. It took: ⏱️ 12 hours 🤖 93 parallel sub-agents 🔄 15k+ model requests 🧠 2.6B tokens processed 💸 Less than $1K in API credits To build a functioning OS from
https://x.com/Google/status/2056789235500466273

We’re bringing generative UI to everyone, free of charge, thanks to Google @Antigravity and the agentic coding capabilities of Gemini 3.5 Flash. Search can build custom visual tools and simulations, tailored to your specific question, on the fly. Under the hood, Search
https://x.com/Google/status/2056795269694423065

Welcome to Gemini 3.5 Flash, our most powerful model to date. It pushes the frontier of intelligence, speed, and cost putting 3.5 Flash in a class of its own. We spent the last 6 months making sure Flash is great for real world use cases. It’s available everywhere now!
https://x.com/OfficialLoganK/status/2056792266514329914

what the actual fuck Gemini 3.5 Flash is 7.46 times more EXPENSIVE than GPT-5.5-xhigh on PencilPuzzleBench (direct ask scores are below gpt-5.2-high)
https://x.com/scaling01/status/2057177354582020362

Google unveils Gemini Spark, a persistent personal AI agent

Anyone understand what Google mean by “”Gemini Spark runs on Gemini 3.5 and uses the Antigravity harness”” – is “”Antigravity”” a generic term they’re using for their agent harnesses now or is their Claw-competitor running the same closed-source Go binary we can download ourselves?
https://x.com/simonw/status/2057115921551098211

Gemini Spark is your 24/7 personal AI agent that helps you navigate your digital life. It transforms Gemini from an assistant that answers your questions, to one that does the work on your behalf, under your direction. #GoogleIO
https://x.com/GeminiApp/status/2056801918018564538

Introducing Gemini Spark ✨ It’s your 24/7 personal AI agent that helps you navigate your digital life, taking action on your behalf, and under your direction. 🧠 It runs on Gemini 3.5 and is built on @Antigravity, so it can perform long-running tasks easily in the background.
https://x.com/Google/status/2056791134295273554

Last month we dropped the Gemini app for macOS. In the coming weeks, we’ll be bringing Gemini Spark to the Gemini desktop app so it can help with tasks like organizing your local files, or extracting PDF data directly into Google Sheets. #GoogleIO
https://x.com/GeminiApp/status/2056802363269329304

Say hello to Gemini Spark, your dedicated agent through the @GeminiApp! It runs on a dedicated virtual machine, can be fully connected to all of your Google info, and is paired with an awesome new UI in the mobile and web app, it looks and feels awesome!
https://x.com/OfficialLoganK/status/2056791703349277101

Is this Google’s answer to OpenClaw, Hermes Agent and Codex?
https://x.com/iScienceLuvr/status/2056792158988816767

SynthID: We’re bringing Gemini Spark to the @GeminiApp for macOS app so it can help with tasks like organizing your local files or extracting PDF data directly into Google Sheets or @Gmail. We’re also bringing powerful new voice understanding technology to your laptop, so you can think
https://x.com/Google/status/2056802434303869118

Google’s Genie world model now simulates real Street View locations

Project Genie is a @GoogleLabs experiment that lets you simulate dynamic worlds you can navigate in real time with Genie, our general-purpose world model. Today, we’re connecting Project Genie to nearly 20 years of Street View data from Google Maps — so you can now build
https://x.com/Google/status/2056850758029464009

Woot! You can now simulate real world places by grounding Genie 3 experiences with Street View imagery. Google sitting on the mother lode of real world data, and is starting to put it to work! Let’s dive into some prompts & locations I tested…
https://x.com/bilawalsidhu/status/2056804315721843024

Google launches Gemini for Science research agents for scientists

New AI Tools for the Future of Science
https://blog.google/innovation-and-ai/technology/research/gemini-for-science-io-2026/

Google unveils Gemini Omni and dual-chip TPUs at I/O 2026

Google I/O 2026: Sundar Pichai’s opening keynote
https://blog.google/innovation-and-ai/sundar-pichai-io-2026/

Google I/O ’26 Keynote – YouTube
https://www.youtube.com/watch?v=wYSncx9zLIU

Google unveils Gemini Omni and dual-chip TPUs at I/O 2028

Sundar Pichai on Agents Replacing Engineers, Google’s Future, AI’s Flip Phone Moment, and More – YouTube
https://www.youtube.com/watch?v=zBOoEpsjWAo

Google expands AI content provenance tools across major products

As generative media becomes more advanced, it’s helpful to know exactly where content comes from — and if it’s been changed. 🕵️ Today, we’re expanding content verification tools across Search, @GeminiApp, @GoogleChrome and @MadeByGoogle. #GoogleIO
https://x.com/Google/status/2056787498676658576

HuggingFace

Simple harness boosts Gemini 3.1 Pro past GPT 5.5 Pro

We released physics-intern: a simple harness for science problems! It gets models like Gemini 3.1 Pro to go from 17.7 -> 31.4, thus beating GPT 5.5 Pro. The physics-intern harness can wrap any model and via dedicated subagent boost the performance of the vanilla reasoning
https://x.com/lvwerra/status/2057476832664953225

Meta

Meta cuts 8,000 jobs while boosting AI spending to $145 billion

Meta’s layoffs starting this week underscore Zuckerberg’s AI reality
https://www.cnbc.com/2026/05/18/metas-layoffs-starting-this-week-underscore-zuckerbergs-ai-reality-.html

Nvidia

NVIDIA’s first custom CPU Vera ships to top AI labs

Vera Arrives: NVIDIA’s First CPU Built for Agents Lands at Top AI Labs | NVIDIA Blog
https://blogs.nvidia.com/blog/vera-cpu-delivery/

OpenAI

OpenAI’s Codex now controls locked Macs remotely from phones

wow i just had codex analyze 3 years worth of text messages… i had it use direct quotes in its analysis and it brought me to tears. if you have mac you can just ask codex to do this. you will need to give it permissions
https://x.com/rileybrown/status/2056171774564553038

Codex anywhere and everywhere, all the time. Now your Mac doesn’t have to be unlocked for Codex to use your computer. From your phone, Codex can securely use apps on your Mac, even when the screen is off and locked.
https://x.com/OpenAIDevs/status/2057536706778378692

Highlights from today’s Codex Thursday launches: 1️⃣ Codex can now securely use apps on your Mac from your phone, even when your Mac is locked and the screen is off.
https://x.com/OpenAI/status/2057617844800794878?s=20

Codex agent autonomously unsubscribes user from 87 mailing lists

I finally used /goal in Codex and I’m absolutely mind blown. I had it look through my last 500 archived emails. Then look for an unsubscribe link and unsubscribe if there is one. It found 87 and it clicked them all. It handled the “are you sure” pages AND flagged 14 that
https://x.com/toddsaunders/status/2056198815825215875

Elon Musk has lost his lawsuit against Sam Altman and OpenAI | Hacker News
https://news.ycombinator.com/item?id=48182754

Jury dismisses all claims in Elon Musk’s lawsuit against OpenAI CEO Sam Altman : NPR
https://www.npr.org/2026/05/18/nx-s1-5822366/musk-altman-openai-jury-verdict-claims-dismissed

Musk loses case against OpenAI | CNN Business
https://www.cnn.com/2026/05/18/tech/openai-musk-lawsuit-verdict

OpenAI model disproves 80-year-old Erdős geometry conjecture autonomously

May 13, 2024: GPT-4o is released and scores below 10% on AIME 2024 (US math olympiad entry questions) Dec 20, 2024: o3-preview saturates AIME 2024 Jul 21, 2025: Gemini DeepThink scores IMO gold May 20, 2026: internal GPT models makes Erdös unit-distance problem breakthrough
https://x.com/scaling01/status/2057246143881609510

An internal OpenAI model has disproved one of the most well-known Erdős problems: the unit distance problem. This is, without doubt, the most impressive achievement of AI in mathematics so far.
https://x.com/thomasfbloom/status/2057177152894771631

An OpenAI model has achieved a major breakthrough in mathematics, by disproving a central conjecture in discrete geometry that was first posed by Paul Erdős in 1946. This is the first time AI has autonomously solved a prominent open problem central to a field of mathematics.
https://x.com/gdb/status/2057182650784452925

An OpenAI model has disproved a central conjecture in discrete geometry | OpenAI
https://openai.com/index/model-disproves-discrete-geometry-conjecture/

just quick napkin math on how long this took (unless i missed where they said): the published CoT summary is 111,145 tokens long. it’s really hard to say how much they summarized, assume 3x-20x reduction in tokens? and i’m assuming this is gpt-5.6 pro, so taking Artifical
https://x.com/willdepue/status/2057213893857165701

The proof came from a general-purpose reasoning model, not a system built specifically to solve math problems or this problem in particular, and represents an important milestone for the math and AI communities.
https://x.com/OpenAI/status/2057176203166171317

Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946. For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids. An OpenAI model has now disproved that
https://x.com/OpenAI/status/2057176201782075690

unfortunately openai didn’t publish the unsummarized chain of thought, but the summary is 125 pages! the model reaches the crucial idea (which it describes as ‘frightening,’ i would love to read the unabridged chain of thought here…) on page 39
https://x.com/voooooogel/status/2057198687307362642

unit-distance-remarks.pdf
https://cdn.openai.com/pdf/74c24085-19b0-4534-9c90-465b8e29ad73/unit-distance-remarks.pdf

Very proud that an OpenAI model disproved Erdős’s longstanding unit distance conjecture, with an elegant and intricate proof that brings sophisticated ideas from algebraic number theory to bear on geometry. For whatever reason, mathematics has been the field most amenable to
https://x.com/markchen90/status/2057517045575774598

ChatGPT adds account-linked personal finance dashboard for Pro users

A new personal finance experience in ChatGPT | OpenAI
https://openai.com/index/personal-finance-chatgpt/

A preview for Pro users: a new personal finance experience in ChatGPT. Pro users in the U.S. can securely connect financial accounts, see where their money is going, and ask questions based on the information they choose to connect. Your full financial picture, now in ChatGPT.
https://x.com/ChatGPTapp/status/2055317612687675545

ChatGPT for personal finance is interesting, but you need to know what questions to ask and have enough experience to fact-check assumptions. It really needs to ship with some pre-built skills to help guide people to productive use cases & give the AI better instructions as well
https://x.com/emollick/status/2055797877713092997

OpenAI’s GPT-5.5 Pro tested on generating original humor

GPT-5.5 Pro faces its hardest academic challenge: to apply the technique from a paper analyzing which word pairs were funny & why to come up with its own It came up with scrotum snorkel, tuba subpoena, waffle coffin, toad commode, diarrhea tiara, banana tribunal & muffin ruffian
https://x.com/emollick/status/2056067610303672422

OpenAI targets September IPO after Musk lawsuit defeat

OpenAI barrels toward IPO that may happen in September | TechCrunch
https://techcrunch.com/2026/05/20/openai-barrels-toward-ipo-that-may-happen-in-september/

Malta becomes first country to offer citizens free ChatGPT Plus

OpenAI and Malta partner to bring ChatGPT Plus to all citizens | OpenAI
https://openai.com/index/malta-chatgpt-plus-partnership/

OpenAI outruns Anthropic by $1 billion in Q1 revenue

Report: OpenAI’s Q1 revenue was $5.7 billion, beating Anthropic – Sherwood News
https://sherwood.news/tech/report-openais-q1-revenue-was-5-7-billion-beating-anthropic/

OpenAI adds SynthID watermarks and public verification tool for images

Advancing content provenance for a safer, more transparent AI ecosystem | OpenAI
https://openai.com/index/advancing-content-provenance/

Robotics

Boston Dynamics’ Atlas robot lifts 50-pound fridge using full body

Atlas hauling a 50 lb mini-fridge – Practiced the maneuver for millions of hours in a virtual environment. – Focused the training policy on full-body engagement rather than just hand-grasping, allowing the robot to leverage its entire frame for the lift.
https://x.com/TheHumanoidHub/status/2056411129837195531

Figure’s humanoid robot autonomously makes a bed in demo video

Figure AI Humanoids Autonomously Make a Bed in a New Video – Business Insider
https://www.businessinsider.com/figure-ai-robots-humanoids-make-a-bed-video-2026-5

Figure’s Adcock: cross-task data transfer learning skyrockets humanoid success rates

Brett Adcock highlights transfer learning as the “”unfair advantage”” of humanoids, where a single NN transfers knowledge across diverse tasks. He cites how the success rate of a fridge-related task jumped from 60% to 90% after including non-fridge data in the data set.
https://x.com/TheHumanoidHub/status/2055355958117986396

Google

Google says AI Search optimization still starts with good SEO

Optimizing your website for generative AI features on Google Search
https://developers.google.com/search/docs/fundamentals/ai-optimization-guide

Full Executive Summaries with Links, Generated by Haiku 4.5

Andrej Karpathy joins Anthropic to lead AI research automation efforts.
Karpathy, a prominent AI researcher who recently left Tesla, has joined Anthropic to establish a new team focused on “autoresearch”—using AI systems to accelerate their own development rather than relying solely on human researchers. The hire signals that frontier AI labs now see automating the research process itself as a critical competitive advantage, moving beyond simply building larger models.

Karpathy was literally hired for RSI he’s starting a new pre-training team at Anthropic that focuses on autoresearch per Axios https://x.com/scaling01/status/2056773883982762114

Personal update: I’ve joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time. https://x.com/karpathy/status/2056753169888334312

I appreciate you sharing this, but I don’t have enough material to create a meaningful summary. The phrase “It’s over Mythos is insane” is too vague for me to determine what actually happened, why it matters, or what evidence supports it.
Could you provide more context? For example: – What is Mythos (a company, model, research paper)? – What specific event or announcement occurred? – Where is this from (news article, social media, report)? With those details, I can deliver the two-line executive summary you’re looking for.

It’s over Mythos is insane https://x.com/scaling01/status/2055362921803211248

Anthropic’s valuation and revenue surge in extraordinary growth phase.
The AI safety company Anthropic tripled its valuation to $900 billion and quintupled annual revenue to $45 billion in just five months, marking one of the fastest scaling trajectories in tech history. This breakneck growth—far outpacing typical enterprise software expansion—signals intense commercial demand for its AI models and reflects the winner-take-most dynamics emerging in the AI market.

Anthropic in a nutshell: Valuation: February 2026: $350B May 2026: $900B Almost 3x in just 3 months. ARR: End of 2025: $9B End of May 2026: $45B Almost 5x in 5 months. Holy. via FT https://x.com/kimmonismus/status/2055222524774846576

Anthropic claims US AI lead could widen to two years by 2028.
Anthropic projects it will create vastly more capable AI systems by 2028 while asserting the US maintains a 12-24 month advantage over Chinese competitors—a significant expansion from the current 5-8 month gap in general and coding capabilities. The claim hinges on Anthropic’s ambitious scaling plans, though it remains unverified and depends on continued US investment and compute access advantages.

Not only is Anthropic saying they will have a “”country of geniuses in a datacenter”” by 2028, but also that the US could be ahead by 12-24 months. Before GPT-5.5 and Claude Mythos chinese labs were ~8 months behind in broader capabilities and ~5 months behind in coding. However, https://x.com/scaling01/status/2055018330365345896

Google integrates AI agents directly into search, launching free this summer
Google is embedding autonomous AI agents into its search platform, starting with information-gathering tools that break down complex questions, assess urgency, and pull real-time web data. The move represents a fundamental shift in how billions of people access information—comparable to Google’s original search revolution—by letting users create multiple AI agents to handle routine research and coding tasks without leaving the search box. The company is rolling out an upgraded search interface powered by its faster Gemini 3.5 Flash model, positioning search itself as an agent operating system rather than a simple query tool.

Agentic coding on the scale of search – this summer for everyone, free of charge #google That’s insane. Google Search will be able to help you build. I mean – I don’t think we yet are realizing what it means on that scale. https://x.com/TheTuringPost/status/2056795871098913209

How information agents work in Search: 1️⃣ Start with a total brain dump of what you want your agent to keep you updated on 2️⃣ The agent will break down your complex question and map out a plan 3️⃣ It will determine the urgency — understanding that you need in-the-moment intel https://x.com/Google/status/2056794675214700764

Introducing our brand new, intelligent Search box — totally reimagined with AI. This is the biggest upgrade to our Search box in 25 years and it’s starting to roll out today. Designed to anticipate your intent, the new Search box helps you formulate your question with AI-powered https://x.com/Google/status/2056793802141044786

Soon, you’ll be able to create and manage multiple AI agents for your many tasks — right in Search ✨ We’re starting with information agents: 🔹These agents intelligently look across everything on the web, including blogs, news sites and social posts, plus real-time data on https://x.com/Google/status/2056794282502054066

This is mind blowing actually. I’m not that easy to impress but this will change everything. This is the second time Google is changing how people search information and what they get as a result. To clarify: Google is integrating the agentic coding capabilities of Gemini 3.5 https://x.com/Kseniase_/status/2056798225378783656

Today we are starting to roll out the biggest upgrade to the Google Search box in over 25 years — now completely reimagined with AI, along with Gemini 3.5 Flash as the new default model for AI mode users globally! https://x.com/OfficialLoganK/status/2056802276124328352

Google releases Gemini Omni, AI that edits video from conversation.
Google launched Gemini Omni, a new AI model that generates and edits videos by combining text, images, audio, and video inputs—available today to paid Gemini subscribers. Unlike previous AI systems, Omni understands real-world physics and can iteratively refine videos through natural conversation, letting users adjust scenes, camera angles, and lighting in multiple turns while maintaining character consistency. The rollout represents Google’s first major step toward “world models” that can create content from any input format.

Gemini Omni is a major leap in world understanding & multimodal editing! It can take photos, video & audio and build entirely new scenes. Over time it’ll be able to handle any input & any output – starting w/ video You can even give it your own videos & iterate on your ideas: https://x.com/demishassabis/status/2056831486251380783

Gemini Omni is coming to the Gemini app for paid subscribers today. It lets you bring your ideas to life using any combination of text, images, and video inputs. Just open up Gemini, attach a video from your camera roll, and change it around. It’s that simple. #GoogleIO https://x.com/GeminiApp/status/2056800579159216202

Gemini Omni is our new AI model that can create anything from any input, starting with video. 🪄 Hear from @DemisHassabis on how you can mix text, audio, and images to generate and edit high-quality videos just by having a conversation. #GoogleIO https://x.com/Google/status/2057180052979409172

Gemini Omni is so fun – insanely great at editing videos! https://x.com/joshwoodward/status/2056827449556845051

Gemini Omni is wild https://x.com/osanseviero/status/2056863263305105424

Gemini Omni was bigger news than Gemini 3.5 Flash https://x.com/scaling01/status/2057143531622334678

Gemini Omni: “”a dramatic reading of Death by Water from the Wasteland by a man eating garlic bread while balanced on a unicycle on a small platform over a churning sea of tomato sauce in which, at the center, sites a meatball with bright blue eyes wearing a top hat”” https://x.com/emollick/status/2056791733619376315

I had early Gemini Omni access: “”sea otter in a pilot’s uniform explains why Spirit Airlines went bankrupt to a river otter who is distracted by their laptop while they are in a hot air balloon over NYC. in the next balloon over, william shakespeare fights a robot made of pizza”” https://x.com/emollick/status/2056788122369712148

Introducing Gemini Omni 🔮…….. Omni is our new model that can create anything from any input — starting with video (think Nano Banana but for video). Available in the Gemini App, Flow, and YouTube, with API support coming soon! https://x.com/OfficialLoganK/status/2056787874260164628

Meet Gemini Omni — our new AI model that can create anything from any input, starting with video. 🪄 #GoogleIO https://x.com/Google/status/2056786395067552140

Nano Banana for video is here! Google has long touted that Gemini is natively multi-modal in & out — but Omni is the first glimpse into the power of that paradigm applied to creation. Toss in a video and do multi-turn edits. Toss in audio and get reactive visuals. It’s kinda https://x.com/bilawalsidhu/status/2056790381514076449

Starting today, Gemini Omni is rolling out to all Google AI Plus, Pro and Ultra subscribers globally at https://t.co/382WL5xSvc and in the app. Soon, we will support more output formats, like image and audio. Give Gemini Omni a try and share your creations in the replies. 👇 https://x.com/GeminiApp/status/2056814117047132301

The Odyssey and the Iliad get so many movie treatments but the sequel, the Roman Aeneid, is entirely ignored. Here is a teaser trailer from one prompt to Gemini Omni. The first pass made all the flags Danish(?) but Omni is capable of editing video, so I asked for their removal. https://x.com/emollick/status/2056855332127711387

The real „wow“ moment is Gemini Omni. A world model towards AGI. It can create anything from any input. This is insane. https://x.com/kimmonismus/status/2056802929957568881

Trying Gemini Omni on real life #google https://x.com/TheTuringPost/status/2057167259877916679

We’re dropping Gemini Omni: our first step towards a model that can create anything from anything – starting with video. It combines Gemini’s intelligence with our generative media systems – representing a leap forward in world understanding, multimodality, and editing 🧵 https://x.com/GoogleDeepMind/status/2056786446636212467

Big improvements in quality and capabilities from the Omni team, excited to see how we translate this to progress for embodied AGI https://x.com/jparkerholder/status/2056789448554062232

By default Omni has a bit of a professional “”look””, but can also create “”normie”” videos with the right prompting. https://x.com/shlomifruchter/status/2056858151987884087

Creating, remixing, and editing a video is easier than ever with Gemini Omni. It offers a fluid, conversational way to create and edit. Just upload a video from your camera roll and ask Gemini to make changes. https://x.com/GeminiApp/status/2057159933934907825

Edit your own videos with Gemini Omni with just a conversation. 🎥 Prompt the changes you want to see to reimagine the action, change the point of view, or adjust the lighting over multiple turns. Every instruction builds on the last, so your characters stay consistent, the https://x.com/Google/status/2056786888930062369

Gemini Omni can create anything from any input, starting with video. 🪄 This means you can combine images, audio, video and text as input and generate high-quality videos. Or use drawings to create in a way that matches your vision. #GoogleIO https://x.com/Google/status/2056786781992071172

Gemini Omni combines an intuitive understanding of physics with Gemini’s real-world knowledge and reasoning. 🌐 Now, the stories and outputs you create won’t just look photorealistic — they’ll behave like the real world. #GoogleIO https://x.com/Google/status/2056786589175677089

Gemini Omni Flash is rolling out starting today. Here’s where you can find it: 🔹 Today: Google AI Plus, Pro and Ultra subscribers globally in the @GeminiApp and @FlowbyGoogle . 🔹Rolling out starting this week, for no cost: @YouTube Shorts and the YouTube Create app. https://x.com/Google/status/2056789307856462061

Gemini Omni Flash: > a recording from a capsule on the london eye, a jerky zoom into something in the distance and then refocusing (with a bit of back and forth) (no timestamp or dialog) Note the world knowledge of London’s landscape, and the way the video is gently moving like https://x.com/fofrAI/status/2056789242274259242

Google’s AI services process 7x more data in a year, now handling 3.2 quadrillion monthly tokens.
Google disclosed a sevenfold surge in AI processing volume over twelve months, alongside doubling its Gemini app user base to over 900 million. The explosive growth in “tokens”—the basic units of data AI systems process—signals unprecedented scale in real-world AI deployment across consumer and developer applications, marking a shift from experimental tools to infrastructure handling vast computational loads.

At last year’s #GoogleIO, we were processing 480 trillion tokens a month across our surfaces. Now, we’re processing over 3.2 *quadrillion* tokens a month. That’s a 7x increase in just a year. These tokens represent problems being solved — by a user or a developer or a https://x.com/Google/status/2056783102085640252

In just a year, @GeminiApp users have more than doubled, surpassing 900 million. #GoogleIO https://x.com/Google/status/2056783643381543253

California launches AI workforce disruption task force to ensure workers share in productivity gains
Governor Newsom signed an executive order directing state agencies to develop new policies protecting workers from AI-driven job displacement while ensuring they benefit from resulting economic gains. The directive calls for modernized job training, enhanced severance standards, workforce ownership models, and real-time tracking of AI-related layoffs across sectors. This marks the first comprehensive state-level effort to address AI’s labor market impacts rather than just regulating the technology itself.

Governor Newsom signs first-of-its-kind executive order to prepare workers and businesses for potential AI disruption | Governor of California https://www.gov.ca.gov/2026/05/21/governor-newsom-signs-first-of-its-kind-executive-order-to-prepare-workers-and-businesses-for-potential-ai-disruption/

Figure AI’s humanoid robots livestreamed eight days of nonstop warehouse work.
Figure AI captured viral attention by livestreaming humanoid robots autonomously sorting packages for over 200 hours without failure, handling nearly 250,000 packages at roughly three seconds per package. The company extended its initial eight-hour demo into a continuous marathon after early success, even staging a human-versus-machine competition where an intern narrowly beat the robots. While the spectacle has drawn comparisons to landmark product demos, experts caution that the demonstration shows narrow, repetitive capability in a controlled setup rather than general-purpose robot ability—and lack of independent verification raises questions about true autonomy.

76,940 packages over 61 hours 18 minutes The robots are averaging roughly 1 package every 2.9 seconds No breaks https://x.com/adcock_brett/status/2055534794667442361

Day 2 is Live: Watch humanoid robots Bob, Frank, and Gary running 24/7. This is fully autonomous running Helix-02 https://x.com/adcock_brett/status/2054970993442169230

Figure is now entering Day 8 of the most boring video we’ve ever posted F.03 has been working fully autonomously, non-stop for a full week – and we’ve livestreamed the entire thing https://x.com/adcock_brett/status/2057145678355460126

Figure just live-streamed 8 hours of fully autonomous, unsupervised work We just huddled internally – and guess what? We’re not stopping here 24/7 LIVESTREAM https://x.com/adcock_brett/status/2054729581391962353

Figure just set a new standard in unedited humanoid demos. They completed 24 hours of autonomous work on a live stream, where three Figure 03 robots took turns processing delivery packages. At a rate of ~21 packages/minute, that is less than 3 seconds per package. https://x.com/TheHumanoidHub/status/2054993551973589304

Figure launched yesterday aiming to hit 8 hours of autonomous work We’ve now crossed 30 hours of continuous operations with no downtime 🔥🔥🔥 We also welcomed Rose to the team – helping push us past 38,000 packages processed https://x.com/adcock_brett/status/2055075231002407417

Figure livestream is back for day 6! Aime got some well deserved sleep, while F.03 has sorted another 17,444 packages https://x.com/adcock_brett/status/2056420115701448915

Livestream Day 4 has begun. We’ve been running F.03 humanoid robots 24/7 – fully autonomous, no breaks, and no downtime https://x.com/adcock_brett/status/2055695727985352722

Silicon Valley’s Latest Binge-Watch Is a Humanoid Warehouse Worker – Business Insider https://www.businessinsider.com/figure-ai-turned-a-humanoid-sorting-packages-must-see-tv-2026-5

The Internet can’t stop watching Figure AI’s humanoid robots handling packages – Ars Technica https://arstechnica.com/ai/2026/05/the-internet-cant-stop-watching-figure-ais-humanoid-robots-handling-packages/

This week, Figure started an 8 hour livestream to demonstrate fully autonomous, nonstop work We’ve now crossed 50 hours of continuous operation with no downtime, sorting over 63,000 packages The robots are still running – we’re going until failure 🔥🔥🔥 https://x.com/adcock_brett/status/2055376836218310926

We’re entering Day 7! We’ve been running close to a week 24/7 and fully autonomous We’ve sorted over 180,000 packages and have been running continuously, without failure, for 144 hours https://x.com/adcock_brett/status/2056786783539692023

Claude agents now execute code within your own secure infrastructure.
Anthropic released self-hosted sandboxes and private network connections for Claude Managed Agents, letting enterprises run AI agents on their own servers or trusted providers like Cloudflare and Vercel instead of Anthropic’s systems. This matters because companies handling sensitive data—financial records, proprietary code, internal databases—can now use AI agents without moving that data to the cloud, maintaining full control over security, logging, and compute resources. Early adopters like Rogo (finance) and Clay (marketing automation) are already building production agents with these guardrails.

New in Claude Managed Agents: self-hosted sandboxes and MCP tunnels | Claude https://claude.com/blog/claude-managed-agents-updates

Anthropic and Gates Foundation commit $200 million to global AI health initiatives.
Anthropic is partnering with the Gates Foundation to deploy its Claude AI system across global health, education, and economic development programs over four years, targeting areas where commercial markets alone won’t reach. The partnership will focus on accelerating vaccine development, improving healthcare decision-making in low-income countries, building educational tools for sub-Saharan Africa and India, and enhancing agricultural productivity for smallholder farmers—representing a significant shift toward using AI for measurable social impact rather than consumer products.

Anthropic forms $200 million partnership with the Gates Foundation \ Anthropic https://www.anthropic.com/news/gates-foundation-partnership

SpaceX and Anthropic ink forty-five billion dollar AI compute partnership.
Anthropic, an AI safety company, committed to paying SpaceX $1.25 billion monthly through May 2029 for access to computing power needed to train and run its Claude AI models—a $45 billion total deal that signals how expensive it has become to build competitive AI systems and reflects intense demand for specialized hardware that few companies can supply at scale.

SpaceX also disclosed exactly how much their deal with Anthropic is worth. They state that they have “”entered Cloud Services Agreements with Anthropic in May 2026 for capacity on COLOSSUS and COLOSSUS II at $1.25B/month through May 2029″” that’s ~$15B/year, and an ~$45B TCV over https://x.com/SemiAnalysis_/status/2057218890288030110

We’re expanding our partnership with @SpaceX, and will be scaling up on GB200 capacity in Colossus 2 throughout June. Appreciate @elonmusk and the team helping us find good homes for the Claudes. https://x.com/nottombrown/status/2057194829986300375

Wow, Anthropic has agreed to pay @SpaceX $1.25 billion per month through May 2029 for AI compute capacity, with capacity ramping in May and June 2026 at a reduced fee. Huge deal https://x.com/SawyerMerritt/status/2057208968741560744?s=20

Rising Treasury yields signal inflation concerns amid AI spending boom.
US borrowing costs have spiked to decade highs as inflation re-accelerates, with markets now pricing in potential interest rate increases rather than cuts—a significant headwind for the capital-intensive AI industry, which is undertaking massive infrastructure investments. This convergence matters because higher rates directly increase the cost of the billions in debt that AI companies rely on to build data centers and train models, potentially slowing the industry’s aggressive expansion plans.

The 10-year US Treasury just hit 4.568%, highest in 10 months. The 30-year is back above 5%. Inflation is re-accelerating, markets have fully priced out any Fed rate cut this year, and some are now betting on a hike. Meanwhile, the AI industry is running the largest debt https://x.com/kimmonismus/status/2055293526125232332

I appreciate you providing material, but this excerpt doesn’t contain sufficient factual information for me to create the executive summary you’ve requested. The text references a report and appendices but doesn’t describe what was actually studied, what risks were assessed, or what key findings emerged.
To produce an accurate two-line summary, I’d need: – The actual findings or conclusions from the Frontier Risk Report – Specific details about what was evaluated or discovered – Why the results matter to business or society Could you share the substantive content from the report itself—such as the main findings, methodology, or key takeaways?

The full Frontier Risk Report includes more than we can describe here, with over 200 pages of appendix materials detailing how the pilot worked, what we asked participants, and what we learned, along with transcripts from our evals. See our website: https://x.com/METR_Evals/status/2056800047258649049

Data center booms lift local incomes but raise living costs sharply.
New research shows AI data centers generate significant economic benefits through construction jobs and increased spending in nearby sectors, boosting overall county income. However, this growth comes with measurable downsides: electricity prices rise and housing becomes more expensive, suggesting benefits concentrate among business owners while everyday residents face higher bills and rents.

“Data centers create economic activity, especially in directly related sectors and during construction, and they are associated with larger county-level income aggregates. They also raise electricity prices and are associated with higher house prices,” https://x.com/emollick/status/2056136463201300572

Artist reveals people can’t tell real Monet from fake AI art.
SHL0MS posted an actual Claude Monet painting while claiming it was AI-generated, drawing hundreds of critical responses about AI art quality—exposing how much public skepticism of AI relies on assumption rather than technical judgment. The experiment highlights a gap between widespread concerns about AI-generated content and people’s actual ability to distinguish it from human creativity, raising questions about how we evaluate and discuss AI’s real capabilities.

An artist just put a mirror to a lot of society’s knee-jerk critiques of AI, using Claude (Monet). SHL0MS, known for his viral experiments, posted an image of a real Monet water lilies painting on X and told people it was AI-generated. Hundreds of replies poured in explaining https://x.com/TheRundownAI/status/2054942919367197112?s=20

Pope Leo XIV launches Vatican AI ethics initiative ahead of major encyclical.
The Vatican announced a new study group on artificial intelligence as Pope Leo XIV prepares to release his first encyclical on the technology, expected within weeks. The encyclical will frame AI through the lens of Catholic social teaching, emphasizing human dignity, labor rights, and peace—drawing a parallel to Pope Leo XIII’s 1891 response to the Industrial Revolution. This positions the Catholic Church as a major institutional voice pushing back against the Trump administration’s deregulatory AI approach, while raising specific concerns about deepfakes, autonomous weapons, and the environmental toll of data centers.

Vatican prepares for Pope Leo XIV’s AI-focused first encyclical | AP News https://apnews.com/article/vatican-artificial-intelligence-pope-musk-nvidia-trump-889c0066f0d5ce784c07abb72b33e24c

AI agents get their own desktop workspace from Google.
Google launched a standalone desktop app designed specifically for AI agents rather than human-computer chat, emphasizing how agents interact with each other and produce work artifacts. This reflects a shift in how companies are building AI tools—moving beyond chatbots toward systems where multiple AI agents collaborate on tasks independently, signaling that agent-to-agent workflows are becoming central to enterprise software design.

Say hello to the new standalone desktop application for Google @Antigravity. 💻 It’s agent-first — focusing on core agent conversations, agent-produced artifacts, and multi-agent orchestration. #GoogleIO https://x.com/Google/status/2056788868092006891

Google launches Gemini agent that auto-digests your email, calendar and tasks
Google has released Daily Brief, an AI agent within its Gemini app that automatically pulls information from your inbox, calendar, and task lists each morning to create a personalized digest highlighting what matters most. The tool aims to save time by condensing scattered information into one skimmable summary, letting users prioritize their day before leaving the breakfast table—a practical example of AI agents handling routine information synthesis rather than creating entirely new capabilities.

Introducing our newest out-of-the-box agent, Daily Brief. It creates a personalized digest in @GeminiApp that’s designed to be your first stop every morning. It synthesizes information from your inbox, your calendar and your tasks to find the most important things for you to be https://x.com/Google/status/2056801159071883342

Daily Brief is a new personalized digest that’s designed to be your first stop every morning. It gathers info from your inbox, calendar, and tasks to prioritize, organize, and suggest the next steps for you in a super concise morning digest that’s built for skimming. #GoogleIO https://x.com/GeminiApp/status/2056800978343764238

Get a head start on your day with Daily Brief. Gemini can now proactively flag what matters most in an easily digestible to-do list, so you’re ready for the day before you even finish breakfast. https://x.com/GeminiApp/status/2057500470147698936

Google launches AI-powered smart glasses this fall with partners
Google is releasing audio glasses from Gentle Monster and Warby Parker that use its Gemini AI assistant to handle tasks like navigation, texting, and photo editing through voice commands—keeping users hands-free and phone-free throughout the day. The glasses represent a shift from smartphones to wearable devices as the primary interface for AI assistance, with display glasses following later. This matters because it signals a fundamental change in how people access AI: embedded in everyday objects rather than pulled from a pocket, designed by fashion brands rather than tech companies alone.

Intelligent eyewear with Gemini is coming this fall https://blog.google/products-and-platforms/platforms/android/android-xr-io-2026/

Google releases Gemini 3.5 Flash for autonomous task execution
Google launched Gemini 3.5 Flash, a new AI model optimized for agents—AI systems that independently execute multi-step workflows—delivering performance comparable to larger flagship models while running four times faster and at less than half the cost. The model outperforms Google’s previous generation on coding and real-world task benchmarks, with enterprise partners like Shopify and Macquarie Bank already using it to automate weeks-long processes. Available free to consumers and via API to developers, 3.5 Flash represents Google’s strategic pivot toward practical, action-oriented AI rather than pure conversation capability.

Gemini 3.5: frontier intelligence with action https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-5/

Gemini 3.5 feels like the start of a new era for Gemini, we spent the last 2.5 years putting the infrastructure, products, team, etc in place (learning lots of lessons along the way). The model is the product, please keep the feedback coming! https://x.com/OfficialLoganK/status/2057104170310902166

Gemini can now connect to even more apps, including @OpenTable, @Canva, and @Instacart. Whether you’re booking a table at a restaurant, creating a flyer, or ordering groceries, Gemini doesn’t just find info, it helps you take action seamlessly with connected apps. https://x.com/GeminiApp/status/2057550225863246236

(deliberately not hyping Gemini 3.5 Flash too much this time. looks like an insane model, but you know how it is with self-reported benchmarks) https://x.com/scaling01/status/2056794370909593987

@demishassabis @antigravity @GeminiApp Are you aware that running AA intelligence index cost almost 2x with 3.5 Flash than it did with 3.1 Pro? https://x.com/giffmana/status/2057155343390494949

⚡ Gemini 3.5 Flash is now available in @code. Give it a try! https://x.com/code/status/2056803208559759447

📣 @GoogleAI’s Gemini 3.5 Flash is now generally available and rolling out in GitHub Copilot. Early testing shows ➡️ It has strong tool use, fast response times, and high cache efficiency ➡️ It is it well-suited for fast, iterative agentic coding workflows Try it out in @code. https://x.com/github/status/2056801675042779279

1/ Today at #GoogleIO, we’re releasing Gemini 3.5, our latest family of models combining frontier intelligence with action. We’re starting by releasing 3.5 Flash, which is built to help you execute complex, long-horizon agentic workflows. Gemini 3.5 Flash is our strongest model https://x.com/JeffDean/status/2056793419033588091

A closer look at Gemini 3.5 Flash by @GoogleDeepMind In the Code Arena: Frontend we see sweeping gains, and a Flash model now surpasses the previous Pro variant. – vs. 3 Flash, a +70 jump overall, large improvements in every subcategory – vs. 3.1 Pro, outperforms it in every https://x.com/arena/status/2056803661859479812

Also had some early access to Gemini 3.5 Flash. Very fast for a flash model and very capable, though not as powerful as a full frontier model. I added it to the gallery or procedurally generated one-shot towns (it made one error that it corrected): https://x.com/emollick/status/2056798490353705380

Artificial Analysis benchmarks were featured in yesterday’s Gemini 3.5 Flash launch Yesterday @GoogleDeepMind released Gemini 3.5 Flash at Google I/O ’26 and our benchmarks were used by @sundarpichai to highlight the model’s leading position on the Intelligence vs. Speed Pareto https://x.com/ArtificialAnlys/status/2057181290412261557

Clearly these people haven’t experienced the magic of Gemini 3.5 Flash Preview on High in the new Antigravity CLI https://x.com/theo/status/2056826014739890204

Gemini 3.5 Flash can detect and reason I asked it to detect and label cars on each lane replacing a custom model + post-processing pipeline with a single API call https://x.com/skalskip92/status/2057502215506473121

Gemini 3.5 Flash delivers fast, consistent performance that rivals other leading models – at a fraction of the price. It can plan and reason across massive codebases and deploy subagents to work in parallel over a long horizon. It outperforms 3.1 Pro on coding and agentic https://x.com/GoogleDeepMind/status/2056787990110994511

Gemini 3.5 Flash delivers sustained frontier-level performance at lightning-quick speeds: Beats 3.1 Pro on coding & agentic benchmarks Terminal-Bench 2.1 (76.2%), GDPval-AA (1656 Elo) and MCP Atlas (83.6%) 4x faster than other frontier models (12x in Antigravity!) SOTA on https://x.com/koraykv/status/2056795667088204234

Gemini 3.5 Flash has landed #9 for Text and Code Arena: Frontend. Code Arena: Frontend evaluates models on agentic frontend coding tasks from real users building apps and websites (HTML and React). Scoring 1507, this is a significant +70 point improvement over Gemini-3 Flash. https://x.com/arena/status/2056793176720195693

Gemini 3.5 Flash is an insane step forward. Google now has to thread the needle and make a great agentic coding product akin to Codex/Claude Code. Does Gemini CLI even exist? Oh it’s now Antigravity CLI? Ugh… Google being Google. https://x.com/zachtratar/status/2056848643580482002

Gemini 3.5 Flash is available now globally to: ✨ Everyone: @GeminiApp and in AI Mode in Google Search ✨ Developers: In Google @Antigravity and Gemini API in @GoogleAIStudio & @Android Studio ✨ Enterprises: In Gemini Enterprise Agent Platform We’re just getting started — https://x.com/Google/status/2056791527314387208

Gemini 3.5 Flash is now GA. Our most capable Flash model, built for agentic execution, coding, and long-horizon tasks. – Outperforms Gemini 3.1 Pro on coding and agentic tasks – 1M token context window with 65k max output tokens – 4x faster output tokens/sec – 4 thinking levels: https://x.com/_philschmid/status/2056794978517750165

Gemini 3.5 Flash is our strongest agentic and coding model yet, outperforming Gemini 3.1 Pro on challenging coding and agentic benchmarks. https://x.com/Google/status/2056788281317306466

Gemini 3.5 Flash is starting to roll out to everyone globally, free of charge. Try it by selecting “3.5 Flash” from the model dropdown menu at https://t.co/382WL5xSvc or in the app. https://x.com/GeminiApp/status/2057140474192994356

Gemini 3.5 Flash is starting to roll out to everyone globally, free of charge. Try it by selecting “3.5 Flash” from the model dropdown menu at https://t.co/382WL5xSvc or in the app. https://x.com/GeminiApp/status/2057237126526517727

Gemini 3.5 Flash Pricing confirmed at $1.5 / $9 per mtoks https://x.com/scaling01/status/2056793465715822720

Gemini 3.5 Flash ranks #1 on the APEX-Agents-AA benchmark, outperforming much larger models a whole size above it. https://x.com/OfficialLoganK/status/2057460544643404125

Gemini 3.5 Flash scores kinda low on the Coding Index due to terrible TerminalBench-Hard scores https://x.com/scaling01/status/2056796392899645919

Google optimized Gemini 3.5 Flash to make it run up to 12x faster (~867 tokens/s) than comparable models in AntiGravity https://x.com/scaling01/status/2056790573961326680

Google’s new Gemini 3.5 Flash is the clear leader on the Intelligence vs Speed Pareto frontier and makes large gains on GDPval-AA (real-world agentic tasks), but is 5x the cost of Gemini 3 Flash @GoogleDeepMind gave us pre-release access to Gemini 3.5 Flash, the latest model in https://x.com/ArtificialAnlys/status/2056795055512596817

GPT-5.5-medium has lower end-to-end latency, uses less tokens and is overall smarter and cheaper than Gemini 3.5 Flash it might genuinely be over for anyone not named OpenAI or Anthropic https://x.com/scaling01/status/2056803273756000721

Insane evals for a Flash model! Gemini 3.5 Flash is really good for its size! https://x.com/kimmonismus/status/2056791681073316071

Introducing Gemini 3.5: our newest family of models combining frontier intelligence with real-world action. The first release is 3.5 Flash, our strongest model yet for agents and coding 🧵 https://x.com/GoogleDeepMind/status/2056787987774816525

Just off stage at #GoogleIO, some highlights from this morning 🧵 Gemini 3.5 Flash is available today for everyone in @antigravity and across our products and APIs. Compared to 3.1 Pro, 3.5 Flash is better across almost all benchmarks with huge progress in coding. It’s also https://x.com/sundarpichai/status/2056796893951426705

Massive updates to @GoogleAIStudio and the Gemini API 🤯 – Gemini 3.5 Flash! – managed agents so you can easily build agentic products with the antigravity harness – native Android app creation right in AI Studio – native workspace integrations – 1 click export to antigravity https://x.com/OfficialLoganK/status/2056840620610830638

Meet Gemini 3.5 Flash — our strongest agentic and coding model yet. It delivers frontier-level performance at 4x the speed of comparable frontier models — often at less than half the cost. Generally available, starting today. 🧵 #GoogleIO https://x.com/Google/status/2056788266872140232

My notes on Gemini 3.5 Flash – 3x the price of Gemini 3 Flash but Google are planning to use it for many of their own products https://x.com/simonw/status/2056867815605625172

Our new Gemini 3.5 Flash is our strongest agentic and coding model yet, delivering frontier-level performance at 4x the speed of comparable frontier models at less than half the cost. We had @TulseeDoshi break down what you need to know about it. https://x.com/Google/status/2057257773868388448

some more Gemini 3.5 Flash benchmarks by Artificial Analysis AI: – the APEX-Agents-AA score is excellent – expected higher on CritPt – reasoning efficiency could also be better, but it kind of depends on what setting they used. if it’s the max then it’s very good – Price/Perf https://x.com/scaling01/status/2056798645983334890

Starting today, Gemini 3.5 Flash is rolling out to everyone globally, free of charge. Try it by selecting “3.5 Flash” from the model dropdown menu at https://t.co/382WL5xSvc or in the app, and let us know how you’re using it in the replies. https://x.com/GeminiApp/status/2056789742910595342

TPUs are insane Gemini 3.5 Flash is running at ~867 tokens/s almost as fast as Kimi-K2.6 on Cerebras custom chips https://x.com/scaling01/status/2056791726677782743

We asked our agents to build a working operating system from scratch using @Antigravity 2.0 and Gemini 3.5 Flash. It took: ⏱️ 12 hours 🤖 93 parallel sub-agents 🔄 15k+ model requests 🧠 2.6B tokens processed 💸 Less than $1K in API credits To build a functioning OS from https://x.com/Google/status/2056789235500466273

We’re bringing generative UI to everyone, free of charge, thanks to Google @Antigravity and the agentic coding capabilities of Gemini 3.5 Flash. Search can build custom visual tools and simulations, tailored to your specific question, on the fly. Under the hood, Search https://x.com/Google/status/2056795269694423065

Welcome to Gemini 3.5 Flash, our most powerful model to date. It pushes the frontier of intelligence, speed, and cost putting 3.5 Flash in a class of its own. We spent the last 6 months making sure Flash is great for real world use cases. It’s available everywhere now! https://x.com/OfficialLoganK/status/2056792266514329914

what the actual fuck Gemini 3.5 Flash is 7.46 times more EXPENSIVE than GPT-5.5-xhigh on PencilPuzzleBench (direct ask scores are below gpt-5.2-high) https://x.com/scaling01/status/2057177354582020362

Google launches Gemini Spark, an always-on personal AI agent.
Google introduced Gemini Spark, a 24/7 AI agent built on its Gemini 3.5 model that autonomously handles tasks like file organization and data extraction across Google services. Unlike traditional chatbots that answer questions, Spark executes actions on users’ behalf—positioning it as Google’s competitor to similar agent products from rivals. The agent runs on dedicated infrastructure (via “Antigravity,” Google’s agent framework) and will roll out across mobile, web, and desktop apps in coming weeks.

Anyone understand what Google mean by “”Gemini Spark runs on Gemini 3.5 and uses the Antigravity harness”” – is “”Antigravity”” a generic term they’re using for their agent harnesses now or is their Claw-competitor running the same closed-source Go binary we can download ourselves? https://x.com/simonw/status/2057115921551098211

Gemini Spark is your 24/7 personal AI agent that helps you navigate your digital life. It transforms Gemini from an assistant that answers your questions, to one that does the work on your behalf, under your direction. #GoogleIO https://x.com/GeminiApp/status/2056801918018564538

Introducing Gemini Spark ✨ It’s your 24/7 personal AI agent that helps you navigate your digital life, taking action on your behalf, and under your direction. 🧠 It runs on Gemini 3.5 and is built on @Antigravity, so it can perform long-running tasks easily in the background. https://x.com/Google/status/2056791134295273554

Last month we dropped the Gemini app for macOS. In the coming weeks, we’ll be bringing Gemini Spark to the Gemini desktop app so it can help with tasks like organizing your local files, or extracting PDF data directly into Google Sheets. #GoogleIO https://x.com/GeminiApp/status/2056802363269329304

Say hello to Gemini Spark, your dedicated agent through the @GeminiApp! It runs on a dedicated virtual machine, can be fully connected to all of your Google info, and is paired with an awesome new UI in the mobile and web app, it looks and feels awesome! https://x.com/OfficialLoganK/status/2056791703349277101

Is this Google’s answer to OpenClaw, Hermes Agent and Codex? https://x.com/iScienceLuvr/status/2056792158988816767

SynthID: We’re bringing Gemini Spark to the @GeminiApp for macOS app so it can help with tasks like organizing your local files or extracting PDF data directly into Google Sheets or @Gmail. We’re also bringing powerful new voice understanding technology to your laptop, so you can think https://x.com/Google/status/2056802434303869118

Google grounds AI world simulator in two decades of Street View data.
Google has connected its Project Genie—a model that generates interactive, navigable 3D environments—to nearly 20 years of Street View imagery, letting users simulate real-world locations in real time. This matters because it demonstrates how companies with vast existing data assets can unlock new capabilities: rather than building synthetic worlds from scratch, Google is leveraging its unmatched repository of real street-level photography to train and ground its generative models in actual geography.

Project Genie is a @GoogleLabs experiment that lets you simulate dynamic worlds you can navigate in real time with Genie, our general-purpose world model. Today, we’re connecting Project Genie to nearly 20 years of Street View data from Google Maps — so you can now build https://x.com/Google/status/2056850758029464009

Woot! You can now simulate real world places by grounding Genie 3 experiences with Street View imagery. Google sitting on the mother lode of real world data, and is starting to put it to work! Let’s dive into some prompts & locations I tested… https://x.com/bilawalsidhu/status/2056804315721843024

Google launches four AI tools designed to speed up scientific discovery and hypothesis testing.
Google has released Gemini for Science, a suite of AI tools built to compress months of manual research into weeks or days—including systems for hypothesis generation, computational experiments, and literature analysis. The tools are designed as “force multipliers” that handle routine synthesis work, freeing researchers to focus on high-impact directions rather than data wrangling. Early enterprise users like BASF and Bayer are already running preview versions, while papers published today in Nature provide validation for the underlying technology.

New AI Tools for the Future of Science https://blog.google/innovation-and-ai/technology/research/gemini-for-science-io-2026/

Google shifts to AI agents that autonomously complete complex tasks across products.
At Google I/O 2026, CEO Sundar Pichai announced the “agentic Gemini era,” marking a transition from AI assistants that respond to queries to AI agents that independently handle multi-step workflows. The company is processing 3.2 quadrillion tokens monthly—a sevenfold increase from last year—and has deployed conversational AI across search, maps, YouTube, and productivity tools like Docs Live, which now accepts voice commands. This shift matters because it signals the industry moving beyond chatbots toward autonomous systems embedded in everyday products, supported by Google’s $180–190 billion annual infrastructure investment and new custom chips optimized for AI workloads.

Google I/O 2026: Sundar Pichai’s opening keynote https://blog.google/innovation-and-ai/sundar-pichai-io-2026/

Google I/O ’26 Keynote – YouTube https://www.youtube.com/watch?v=wYSncx9zLIU

Google’s CEO says AI agents will replace some engineering jobs within five years.
Sundar Pichai told investors that autonomous AI agents—software that can independently complete complex tasks—will increasingly handle work currently done by human engineers, marking a shift toward AI systems that operate without constant human direction. This matters because it signals Silicon Valley’s largest companies see near-term job displacement in white-collar fields, not just automation of routine tasks. Pichai’s timeline of five years is notably concrete compared to earlier industry claims about AI’s distant impact.

Sundar Pichai on Agents Replacing Engineers, Google’s Future, AI’s Flip Phone Moment, and More – YouTube https://www.youtube.com/watch?v=zBOoEpsjWAo

Google expands AI tools to detect AI-generated and altered images.
Google is rolling out content verification features across its search engine, Gemini chatbot, Chrome browser, and hardware products to help users identify AI-generated or manipulated images. This matters because as synthetic media becomes harder to distinguish from authentic content, verification tools are becoming essential infrastructure for combating misinformation and maintaining trust in visual information online.

As generative media becomes more advanced, it’s helpful to know exactly where content comes from — and if it’s been changed. 🕵️ Today, we’re expanding content verification tools across Search, @GeminiApp, @GoogleChrome and @MadeByGoogle. #GoogleIO https://x.com/Google/status/2056787498676658576

Physics-intern harness doubles AI performance on science problems.
Google released a lightweight software tool that dramatically improves how AI models solve physics and science questions—boosting Gemini 3.1 Pro’s accuracy from 17.7% to 31.4%, outperforming OpenAI’s GPT-5.5 Pro. The approach works by adding a specialized sub-agent that enhances reasoning without requiring model retraining, suggesting that clever prompting and task structure can matter as much as raw model size.

We released physics-intern: a simple harness for science problems! It gets models like Gemini 3.1 Pro to go from 17.7 -> 31.4, thus beating GPT 5.5 Pro. The physics-intern harness can wrap any model and via dedicated subagent boost the performance of the vanilla reasoning https://x.com/lvwerra/status/2057476832664953225

Meta cuts 8,000 jobs while betting billions on artificial intelligence.
Meta is laying off 10% of its workforce this week—8,000 employees—while simultaneously increasing AI spending by up to $10 billion, signaling a shift toward automation rather than human labor. Unlike earlier layoffs framed as correcting pandemic hiring mistakes, CEO Zuckerberg offered no apology, and employees report growing dread as more cuts are planned for August and fall. The move reflects a broader tech industry pattern where AI investments and job cuts are happening in tandem, with Wall Street rewarding companies that prioritize automation, even as internal morale and employee ratings at Meta have plummeted.

Meta’s layoffs starting this week underscore Zuckerberg’s AI reality https://www.cnbc.com/2026/05/18/metas-layoffs-starting-this-week-underscore-zuckerbergs-ai-reality-.html

NVIDIA launches Vera, its first custom processor designed specifically for AI agents.
NVIDIA released Vera, a CPU built from the ground up to run AI agent software rather than general computing tasks, now deployed at leading AI research labs. This marks a shift from adapting existing chips toward purpose-built hardware for the emerging category of autonomous AI systems that can plan and execute multiple steps independently. The move signals industry confidence that AI agents—software that acts on its own without constant human direction—are becoming central to AI development rather than experimental.

Vera Arrives: NVIDIA’s First CPU Built for Agents Lands at Top AI Labs | NVIDIA Blog https://blogs.nvidia.com/blog/vera-cpu-delivery/

AI gains persistent access to Mac computers from phones, raising privacy concerns.
Codex, an AI system, can now analyze personal files and control Mac applications remotely—even when the computer is locked—if granted permission. While the capability enables compelling use cases like automated text message analysis, the always-on remote access model raises significant questions about data privacy, consent scope, and what happens when AI systems have continuous access to sensitive personal information without explicit user oversight.

wow i just had codex analyze 3 years worth of text messages… i had it use direct quotes in its analysis and it brought me to tears. if you have mac you can just ask codex to do this. you will need to give it permissions https://x.com/rileybrown/status/2056171774564553038

Codex anywhere and everywhere, all the time. Now your Mac doesn’t have to be unlocked for Codex to use your computer. From your phone, Codex can securely use apps on your Mac, even when the screen is off and locked. https://x.com/OpenAIDevs/status/2057536706778378692

Highlights from today’s Codex Thursday launches: 1️⃣ Codex can now securely use apps on your Mac from your phone, even when your Mac is locked and the screen is off. https://x.com/OpenAI/status/2057617844800794878?s=20

Jury dismisses Musk’s lawsuit against OpenAI over statute of limitations.
A California jury ruled that Elon Musk waited too long to sue OpenAI’s Sam Altman and Greg Brockman, finding his 2024 lawsuit fell outside the three-year statute of limitations—a procedural decision that meant the court never evaluated whether Altman and Brockman actually violated their nonprofit mission by creating a for-profit structure. Musk, who co-founded and funded OpenAI with $38 million before leaving in 2018, had sought over $130 billion in damages and removal of the company’s leaders, but the jury determined he knew about the contested conduct as early as 2021. The verdict removes a significant legal threat to OpenAI ahead of a potential blockbuster IPO, though Musk has indicated he plans to appeal.

I finally used /goal in Codex and I’m absolutely mind blown. I had it look through my last 500 archived emails. Then look for an unsubscribe link and unsubscribe if there is one. It found 87 and it clicked them all. It handled the “are you sure” pages AND flagged 14 that https://x.com/toddsaunders/status/2056198815825215875

Elon Musk has lost his lawsuit against Sam Altman and OpenAI | Hacker News https://news.ycombinator.com/item?id=48182754

Jury dismisses all claims in Elon Musk’s lawsuit against OpenAI CEO Sam Altman : NPR https://www.npr.org/2026/05/18/nx-s1-5822366/musk-altman-openai-jury-verdict-claims-dismissed

Musk loses case against OpenAI | CNN Business https://www.cnn.com/2026/05/18/tech/openai-musk-lawsuit-verdict

OpenAI model solves eighty-year-old geometry problem autonomously.
An internal OpenAI reasoning model has disproven a central conjecture in discrete geometry that mathematician Paul Erdős posed in 1946—the first time AI has autonomously solved a prominent open problem in mathematics. The model discovered that square-grid constructions, long believed optimal, can be beaten through sophisticated techniques from algebraic number theory, with the proof verified by leading mathematicians including Fields Medalist Tim Gowers. This breakthrough demonstrates AI systems can generate original mathematical insights, not merely assist human researchers.

May 13, 2024: GPT-4o is released and scores below 10% on AIME 2024 (US math olympiad entry questions) Dec 20, 2024: o3-preview saturates AIME 2024 Jul 21, 2025: Gemini DeepThink scores IMO gold May 20, 2026: internal GPT models makes Erdös unit-distance problem breakthrough https://x.com/scaling01/status/2057246143881609510

An internal OpenAI model has disproved one of the most well-known Erdős problems: the unit distance problem. This is, without doubt, the most impressive achievement of AI in mathematics so far. https://x.com/thomasfbloom/status/2057177152894771631

An OpenAI model has achieved a major breakthrough in mathematics, by disproving a central conjecture in discrete geometry that was first posed by Paul Erdős in 1946. This is the first time AI has autonomously solved a prominent open problem central to a field of mathematics. https://x.com/gdb/status/2057182650784452925

An OpenAI model has disproved a central conjecture in discrete geometry | OpenAI https://openai.com/index/model-disproves-discrete-geometry-conjecture/

just quick napkin math on how long this took (unless i missed where they said): the published CoT summary is 111,145 tokens long. it’s really hard to say how much they summarized, assume 3x-20x reduction in tokens? and i’m assuming this is gpt-5.6 pro, so taking Artifical https://x.com/willdepue/status/2057213893857165701

The proof came from a general-purpose reasoning model, not a system built specifically to solve math problems or this problem in particular, and represents an important milestone for the math and AI communities. https://x.com/OpenAI/status/2057176203166171317

Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946. For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids. An OpenAI model has now disproved that https://x.com/OpenAI/status/2057176201782075690

unfortunately openai didn’t publish the unsummarized chain of thought, but the summary is 125 pages! the model reaches the crucial idea (which it describes as ‘frightening,’ i would love to read the unabridged chain of thought here…) on page 39 https://x.com/voooooogel/status/2057198687307362642

unit-distance-remarks.pdf https://cdn.openai.com/pdf/74c24085-19b0-4534-9c90-465b8e29ad73/unit-distance-remarks.pdf

Very proud that an OpenAI model disproved Erdős’s longstanding unit distance conjecture, with an elegant and intricate proof that brings sophisticated ideas from algebraic number theory to bear on geometry. For whatever reason, mathematics has been the field most amenable to https://x.com/markchen90/status/2057517045575774598

ChatGPT launches connected finance dashboard for paid U.S. users
OpenAI rolled out a personal finance feature for ChatGPT Pro subscribers that links to over 12,000 financial institutions, allowing users to see spending patterns and ask the AI context-aware questions about budgeting and financial decisions. The move leverages ChatGPT’s 200+ million monthly users already seeking financial guidance, with Intuit partnership support coming soon, though OpenAI cautions it’s not a replacement for professional advice. Critics note early versions require users to know which questions to ask and validate the AI’s reasoning, suggesting the feature could benefit from built-in guidance on optimal use cases.

A new personal finance experience in ChatGPT | OpenAI https://openai.com/index/personal-finance-chatgpt/

A preview for Pro users: a new personal finance experience in ChatGPT. Pro users in the U.S. can securely connect financial accounts, see where their money is going, and ask questions based on the information they choose to connect. Your full financial picture, now in ChatGPT. https://x.com/ChatGPTapp/status/2055317612687675545

ChatGPT for personal finance is interesting, but you need to know what questions to ask and have enough experience to fact-check assumptions. It really needs to ship with some pre-built skills to help guide people to productive use cases & give the AI better instructions as well https://x.com/emollick/status/2055797877713092997

AI’s humor generation stumbles on understanding what makes jokes work.
OpenAI’s latest model was given a linguistic framework for analyzing funny word pairs, then asked to create original humor. Instead of applying the underlying principles, it produced nonsensical rhyming pairs like “scrotum snorkel” and “diarrhea tiara”—suggesting the AI memorized surface patterns rather than grasping why incongruity and surprise actually make people laugh. This reveals a real limitation: even advanced systems struggle to move beyond pattern-matching to genuine creative reasoning.

GPT-5.5 Pro faces its hardest academic challenge: to apply the technique from a paper analyzing which word pairs were funny & why to come up with its own It came up with scrotum snorkel, tuba subpoena, waffle coffin, toad commode, diarrhea tiara, banana tribunal & muffin ruffian https://x.com/emollick/status/2056067610303672422

OpenAI plans September IPO after Musk lawsuit dismissal clears path
OpenAI aims to go public as soon as September after a court rejected Elon Musk’s lawsuit challenging the company’s structure, with Goldman Sachs and Morgan Stanley leading the effort. The timing sets up a financial rivalry between Musk-backed competitors—SpaceX, now owner of AI startup xAI, is also preparing its own IPO filing. The move marks a major milestone for the ChatGPT maker and resolves a legal threat that could have complicated its path to public markets.

OpenAI barrels toward IPO that may happen in September | TechCrunch https://techcrunch.com/2026/05/20/openai-barrels-toward-ipo-that-may-happen-in-september/

Malta becomes first nation to offer free ChatGPT Plus to all citizens.
OpenAI and Malta announced a partnership providing free access to ChatGPT Plus paired with an AI literacy course for all Maltese citizens, starting May 2026. The initiative aims to democratize access to advanced AI tools while building practical skills, positioning Malta as a test case for how governments can adopt AI at scale beyond early adopters and tech-savvy users.

OpenAI and Malta partner to bring ChatGPT Plus to all citizens | OpenAI https://openai.com/index/malta-chatgpt-plus-partnership/

OpenAI’s Q1 revenue of $5.7 billion edges out Anthropic’s $4.8 billion in major milestone quarter.
OpenAI maintained a $1 billion revenue lead over rival Anthropic in the first quarter of 2026, though Anthropic is projected to dramatically close the gap in Q2 with revenue potentially reaching $10.9 billion. Despite OpenAI’s current revenue advantage, Anthropic’s latest fundraising round values it at up to $950 billion—surpassing OpenAI’s $850 billion valuation—signaling investor confidence in Anthropic’s growth trajectory as both companies race toward expected IPOs this year.

Report: OpenAI’s Q1 revenue was $5.7 billion, beating Anthropic – Sherwood News https://sherwood.news/tech/report-openais-q1-revenue-was-5-7-billion-beating-anthropic/

OpenAI launches tools to verify AI-generated images and detect watermarks.
OpenAI is releasing a public verification tool that checks whether images were generated by its systems using invisible watermarks (SynthID) and metadata standards (Content Credentials), addressing growing concern about distinguishing AI-created content from authentic media. The company is adopting industry-standard C2PA conformance to ensure provenance information survives when content is shared across platforms, though the tool acknowledges that no detection method is foolproof and cannot make definitive conclusions when signals are absent.

Advancing content provenance for a safer, more transparent AI ecosystem | OpenAI https://openai.com/index/advancing-content-provenance/

Atlas robot learns to carry heavy objects using full-body coordination.
Boston Dynamics’ Atlas humanoid robot successfully lifted a 50-pound mini-fridge by training in simulation millions of times, emphasizing whole-body mechanics rather than isolated hand strength. This demonstrates progress in teaching robots practical physical tasks through virtual practice, a key challenge in making robots useful for real-world work.

Atlas hauling a 50 lb mini-fridge – Practiced the maneuver for millions of hours in a virtual environment. – Focused the training policy on full-body engagement rather than just hand-grasping, allowing the robot to leverage its entire frame for the lift. https://x.com/TheHumanoidHub/status/2056411129837195531

Figure AI’s humanoid robots autonomously complete complex household tasks.
Figure AI released video evidence of humanoid robots independently making a bed—a significant milestone because it demonstrates machines can execute multi-step physical tasks without human intervention or pre-programmed instructions. This moves beyond simple, repetitive factory work into unstructured home environments where robots must adapt to variable conditions, suggesting practical deployment of humanoids in domestic settings may be closer than previously expected.

Figure AI Humanoids Autonomously Make a Bed in a New Video – Business Insider https://www.businessinsider.com/figure-ai-robots-humanoids-make-a-bed-video-2026-5

Humanoid robots gain efficiency edge by reusing learned skills across tasks.
A single AI model trained on diverse data can dramatically improve performance on specific jobs—one robot’s success rate on a fridge task jumped from 60% to 90% after learning from unrelated scenarios. This “transfer learning” approach lets robots solve new problems faster without starting from scratch, giving companies with diverse training data a significant competitive advantage in robotics deployment.

Brett Adcock highlights transfer learning as the “”unfair advantage”” of humanoids, where a single NN transfers knowledge across diverse tasks. He cites how the success rate of a fridge-related task jumped from 60% to 90% after including non-fridge data in the data set. https://x.com/TheHumanoidHub/status/2055355958117986396

Google tells website owners that SEO fundamentals still matter for AI search results.
Google released official guidance advising website owners that traditional search engine optimization (SEO) remains the foundation for visibility in AI-powered search features like AI Overviews. The company emphasizes that its generative AI search uses the same ranking systems as conventional search, meaning websites should focus on creating unique, valuable content rather than chasing new “AI optimization” buzzwords. This matters because it establishes clear rules amid confusion about whether AI search requires entirely different strategies, and it signals Google’s commitment to keeping human expertise and original reporting valuable in an AI-driven search landscape.