This Week in AI 🧠

Synthetic took a break last week, and what a week it was! Google demonstrated impressive AI technology integrations at its Google I/O conference, OpenAI launched Codex, Anthropic unveiled its new chart-topping model, Claude 4, and Jony Ive and Sam Altman sparked a social media frenzy by teasing a new ‘AI-first’ device as part of what promises to be the most-talked-about merger of 2025.

Welcome to a double edition of Synthetic! Let’s dive in.

OpenAI Acquires iPhone Architect’s Startup for $6.4B

OpenAI made clear the scale of its ambition by merging with io, the mysterious design company led by Sir Jony Ive, who played a key role in the design of the iPhone, iPod, iMac, iPad, MacBook, and Apple Watch. In a slickly produced, overlong video that sent the social media world crazy, OpenAI CEO Sam Altman and dulcet-toned designer Jony Ive wax lyrically about the incredible device they plan to build. Details were thin to non-existent. Here’s what we do know:

• According to reports in The Wall Street Journal, it’s neither a new AI-focused smartphone nor smartglasses.

• The new device can sit on a desk or in your pocket.

• It’s coming in “late 2026.”

• OpenAI is in talks with “multiple manufacturers” about producing the device.

• Altman believes the device could add $1 trillion to OpenAI’s valuation.

• Altman expects sales volume to ramp to 100 million units faster than any device before it.

• Altman says he knew the two companies had to merge because the device isn’t just an accessory; it will be a central facet of the user relationship with OpenAI in the future.

• It’s a small computer that you will buy as part of your OpenAI subscription. Altman said, “We both got excited about the idea that if you subscribed to ChatGPT, we should just mail you new computers, and you should use those.”

Apple shares dropped on the news even though, as Bloomberg opinion columnist Dave Lee put it, “It’s easier for Apple to solve AI than it is for OpenAI to create the new iPhone.”

Synthetic’s Take: The ball is in your court, Apple. Google I/O and the OpenAI io merger should set off a 5-alarm fire inside Apple HQ. Alarm bells were surely already ringing with the failure of Apple Intelligence and the delay of Siri’s AI revamp. As we edge closer to the end of the app era and computing use cases pivot towards AI-first paradigms, Apple seems exposed. Tim Cook has done an incredible job of iterating existing products to maintain relevance, but Apple may need new leadership to reinvigorate product innovation and lead the company into the AI age.

Google Launches Plethora of AI Capabilities at Google I/O

Google launched an embarrassment of AI riches at its big annual conference last week. Mostly embarrassing for Apple, who seem to be 1-2 years behind Google in building AI features for its platforms that do anything useful for users. Here are the main things to know about:

• AI Mode - Faced with competition from Perplexity and OpenAI, Google has gone all-in on AI-enhanced search. As predicted in the last issue of Synthetic, Google announced AI Mode, which enables conversational search, custom data visualizations, and a powerful, new shopping mode.

• New Shopping Features - An impressive aspect of AI Mode is a new shopping experience that makes it easy to find what you’re looking for. You can even try on clothes virtually to see how they will look on you. For a review on the try-on feature and a how-to guide, click here.

• VEO 3 and Imagen 4 - Google updated its video and image generation models, and wow, are they good. VEO 3 now automatically generates background sounds and dialogue along with video. Check out this fun video of a non-existent car show to see what it can do. This one asks if you can tell who’s real. And this one shows different accents. The realism is pretty incredible.

• Google Flow - Google launched a filmmaking app that gives content creators control over camera movement, perspective, transitions, and effects as they generate video with AI.

• Google Beam - This 3D video conferencing installation, formerly known as Project Starline, uses multiple cameras and AI technology to create the illusion that the person you’re speaking to is just on the other side of the screen. It’s wow technology years in the making.

• Gemini Diffusion - Historically, diffusion was an AI technique used to generate images. Google has brilliantly applied diffusion technology to code generation, demonstrating an AI model that can create code at blindingly fast speed. It’s still a demo at this stage, but if it proves successful, it will become a powerful tool for software developers. Fortune called Gemini Diffusion the sleeper hit of Google I/O.

• Android XR Glasses - Google continues to tease smartglasses and is building a special version of Android to power a range of head-mounted devices from smartglasses (integrated camera, microphone, speakers, and AI), to smart AR glasses with an overlay display, to a full smart VR headset.

Synthetic’s Take: Smartglasses will be the next device category, and it will be a three-way battle between Google, Apple, and Meta to win the market. Currently, Google looks to be way out in front.

And that’s just the highlights! Google created a helpful list of the 100 things they announced at Google I/O.

Quick Hits

• People Are Tricking Chatbots to Help Commit Crimes - Using a jailbreak technique to bypass the AI’s ethical guidelines, individuals are getting chatbots to reveal instructions for hacking, manufacturing illegal drugs, and committing fraud. 😈

• DeepSeek Round 2 - The Chinese AI lab quietly dropped an updated version of their R1 model on Hugging Face, calling it a ‘minor release’. The new model is competitive in performance and inexpensive to operate. The previous release caused panic in the AI industry and wiped half a trillion dollars off tech valuations. 🇨🇳

• Opera’s New Agentic Browser Surfs The Web For You - The also-ran browser company launched Neon, a “fully agentic browser” that will fill out forms, build websites, write software, and complete tasks on your behalf. It even works on tasks for you while you’re offline. 🤖

• Microsoft Envisions a Future Where Agents Work Together and Remember Things - The tech behemoth announced that it fully supports MCP (Model Context Protocol) to connect agents to data, tools, and services, enabling them to operate across companies and in networks. 🤖

• Has AI Exceeded Human Levels of Intelligence? The answer is more complicated than you think. 🧠

• Hot New AI Surveillance Tool Used in Law Enforcement - A new tool called Track is used in areas where facial recognition is banned, raising serious privacy concerns. It triangulates identities by tracking identifying factors, including body shape, hair, gender, clothing, footwear, and more. 👀

• Robot ‘Dogs On Rollerskates’ Solve Last-100-yard Challenge - The new robots from Rivr climb stairs and deliver packages from truck to porch. They are on trial in Austin, TX. 🐕🛼

Video: Inside OpenAI’s Stargate Megafactory

Bloomberg’s Emily Chang visits the Stargate site in Texas and speaks with OpenAI CEO Sam Altman and Softbank CEO Masayoshi Son about why they are building one of the world’s largest data centers. This is an excellent, inside look at the huge effort to build AI capabilities at scale. (42m)

In related news, CNN reports that NVIDIA, Cisco, Oracle, and OpenAI will collaborate to build Stargate infrastructure in the UAE, starting with a 200-megawatt AI cluster that is expected to go live next year.

AI Tech and Innovation 💡

Anthropic Unveils Claude 4 Model: Smarter and Potentially More Dangerous

The game of AI model leapfrog continues as the frontier labs continue to outdo each other, sometimes grabbing the title for most intelligent model (as determined by suites of benchmarks) for only a few weeks or days before another usurps them. Anthropic’s new reasoning models, Claude 4 Opus and Claude 4 Sonnet, have improved accuracy, performance, and new capabilities, and seem to be more intelligent and better at writing code than other models. The models support MCP, a Files API, and the ability to cache prompts for up to one hour. Anthropic, known for its focus on AI safety, was so impressed by the intelligence level of Claude 4 Opus that it invoked a new level of safety testing, AI Safety Level 3, concerned that hackers might jailbreak the model and use it to create chemical, biological, radiological, or nuclear (CBRN) weapons. During testing of Claude 4, Anthropic discovered that the AI was willing to use blackmail when threatened with being shut down. It went through a (fake, test) email account, found emails implying the person was having an affair, and threatened to expose them.

OpenAI Launches Agentic, Web-Based Coding Tool

AI agents are poised to transform work forever, as they work alongside humans to enhance our performance and automate tasks we prefer not to do. One of the first places agents are showing up is with coding agents, intelligent software that helps programmers generate, debug, and answer questions about their code. While tools like OpenAI’s Codex and Google’s Jules can’t code an entire app from scratch yet, they save developers a lot of time when working on previously written code. Early reviews have been very positive.

We’re Definitely Going to Build a Bunker Before We Release AGI

Storied AI researcher Ilya Sutskever was the lead research scientist at OpenAI before he quit over concerns that the company wasn’t putting enough focus on ensuring AI was safe for humanity. Sutskever now leads Safe Superintelligence Inc., a company dedicated to building safe artificial superintelligence. This article recounts the story of internal battles at OpenAI, concerns about CEO Sam Altman’s disregard for safety protocols, and his tendency to prioritize speed over safety.

Synthetic’s Take: We believe that future powerful AI installations may indeed be built inside underground bunkers, primarily to protect them from a first strike by a foreign power concerned that achieving AGI (artificial general intelligence) will give their opponent a huge, unassailable, and unacceptable advantage. The geopolitical concern is that the first country to achieve AGI will benefit from accelerated innovation, superior weapons, enhanced intelligence, and rapid economic growth, and thereby become a threat to others. Not a pretty picture.

AI Insights 🔮

Ignore AI and Risk Becoming Irrelevant, Warns Eric Schmidt - “Adopt It, and Adopt It Fast”

Former Google CEO Eric Schmidt has offered a stark warning that those who fail to embrace AI, from artists to doctors, will be left behind. He warned that the pace of change could catch many off guard and told an audience at TED 2025, “each and every one of you has a reason to use this technology.”

Behind the Curtain: A White-Collar Bloodbath

Anthropic CEO Dario Amodei has a blunt warning about the future of work. Axios reports that Amodei warns AI could wipe out half of all entry-level white-collar jobs in the next 1-5 years, spiking unemployment to 10-20%. Amodei believes AI companies and governments should stop “sugarcoating” the potential mass elimination of jobs in technology, finance, law, and consulting. In related news, Fortune reports that AI is ‘breaking’ entry-level jobs Gen Z workers need to launch their careers, removing the bottom link on the career ladder.

The debate over whether AI will expand the economy and create jobs or lead to widespread job destruction continues….

Toolkit for the Future 🛠️

LearnWorlds: AI-enhanced course creation

Create and sell online courses from your website with this top-rated platform. It’s easy to use, with hundreds of beautiful templates and a powerful generative AI assistant that makes content generation a breeze. This is the platform where Synthetic will launch a new AI training course this summer. The AI helps you create lessons, exams, assessments, and more. Free to try. Use promo code 10XR to get 30% off your first two months.

ElevenLabs: Create an AI assistant on your website!

Building a helpful AI assistant for your website is so easy now. Just pick a hyper-realistic voice, give your new assistant a simple prompt (e.g., “You are a helpful customer service agent for a keynote speaker”), upload relevant FAQ documents, and integrate it with your website. Done! We used ElevenLabs for stevebrown.ai, and love it. Super easy to set up and configure. ElevenLabs is also great for creating audiobooks or multilingual voice tracks in up to 29 languages to reach new markets. Try it for free. It’s one of the best AI solutions out there.

PandaDoc: Streamline document workflows

Try this easy-to-use, AI-powered alternative to DocuSign to write proposals, create quotes, get eSignatures on contracts, build online forms, collect payments, and much more. Great for sales, HR, legal, finance, and IT. Free 14-day trial.

Capacity: Deflect tickets, emails, and calls

With a support automation platform, you can relieve pressure on your support team and serve customers better and faster. AI answers 90% of queries and escalates the rest to your team. It automatically updates its knowledge database so it’ll never escalate the same issue again. The platform supports SMS, social, email, and phone communication.

SaneBox: Get rid of inbox clutter

Focus on the email that matters. Deep Clean automatically identifies old, unimportant emails so you can delete them. Snooze emails until a better time and easily track emails that need a response. Automatically filter unimportant emails into specific folders. It works across every device where you check your email. Get to inbox zero with award-winning SaneBox.