Reid Hoffman's Recent LinkedIn Posts

What if most (or all) of the activity in the modern economy was conducted by agents? Sean Neville, Co-Founder of Catena and Circle, is building for that world. This week on Possible, he describes a new economic architecture shaped by the AI revolution.

Sean Neville, the Co-Founder of Catena and Circle, believes all economic transactions should be executed by AI. He argues that agentic commerce won't just be the fastest way to transact but also the safest and smartest. One of the most interesting points Sean makes: Individuals in the future will have access to elite, investment-grade AI advisors –– resources that until now have only been reserved for large institutions and high net worth individuals.

Ivan Zhao of Notion has a good line about AI: agency is like muscle. It can be built, but it can also atrophy. That feels like a useful frame for this moment. A lot of the debate around AI is stuck between utopianism and panic. But one of the more grounded questions is “what happens to people when a machine starts doing more of the cognitive work that used to keep certain human muscles strong?” We’ve seen versions of this before, throughout history. The printing press weakened some forms of memory and oral tradition. Something was indeed lost, but so much was gained, too. The printing press helped create the conditions for mass literacy, broader access to knowledge, and the Enlightenment. Those who changed with the times, and continued to train their agency using new tools succeeded, while others who didn’t fell behind. That’s because agency is not just intelligence (which can be thought of as raw horsepower). It’s the habit of acting, deciding, and pushing forward. Like a muscle, agency weakens when it isn’t used. The question we must all ask is how to use AI in a way that makes our agency muscle stronger. It can help you practice, or pressure test your ideas. It can even help you accomplish new things you would’ve never thought possible (e.g. building an app from scratch). In that version, AI is not replacing agency, it is increasing the number of exercises you can get in to train it.

Despite this week’s positive overall job print, last month was the worst January for job-cut announcements since the Great Recession. U.S. companies announced over 100,000 cuts. Hiring intentions dropped to their lowest level since 2009. When numbers like that hit the headlines, it’s natural to look for a villain. And in 2026, there’s an easy one: AI. But that’s not the whole story. This week on Possible, we're talking about how to cut through the noise and make sense of all the jobs news:

The New York Times recently published a blind quiz, challenging readers to distinguish between human and AI-generated writing, and the results sparked an uproar. What does it mean if AI can produce writing that people not only mistake for human work, but sometimes even prefer? Will that make human writers less essential? Will it strip away some of what makes writing beautiful in the first place? Here’s how to interpret the quiz... and how we can navigate the transition thoughtfully:

Excited to join Eric Ries to chat about make-or-break moments for companies, and how to navigate them. Join us on Wednesday!

The simple story about AI adoption within larger organizations is that it's all about saving time. That's not necessarily the whole story, though. This week on Possible, Aria and I talk about why using AI a lot can make you more productive *and* leave you with more work:

There are 18,000 human diseases, yet  14,000 of them don't have a single approved treatment. AI might be able to change that, today. The way we discover and deliver treatments is slow, expensive, and siloed. The consequence is that potential treatments that might help in “unrelated” diseases rarely become hypotheses anyone seriously tests, because no one is set up to connect those dots. David Fajgenbaum, MD, MBA, MSc, and Every Cure are building AI tools to change that. AI is good at sifting through huge volumes of scientific and clinical information and connecting findings that sit in different corners of the research world. That’s why Every Cure is building a giant database to compare all ~4,000 FDA-approved drugs against all ~18,000 diseases, then prioritize which ones look most worth testing. The goal is to narrow the field and open the data so researchers and clinicians can move faster on the best candidates. You can also see the broader implications of Dr. Fajgenbaum's work. If you can surface links between medicines and diseases that don’t sit in the same clinical “lane,” you can start to reduce the fragmentation that defines modern healthcare. Hopefully, that body of work leads to more humans getting access to better healthcare more quickly. And in a moment when a lot of AI commentary defaults to dread, projects like Every Cure are a needed counterweight. A clear story about how AI tools, even in their infancy, can provide results for humanity.

We ask every guest on Possible: “Where do you see progress outside of your industry that is inspiring?” The majority of our guests have the same answer: medicine. With breakthroughs in AI, and in the world of molecules, we stand at the precipice of a new world of medicine:

Researchers at King's College put frontier AI models into 21 simulated nuclear crises. In 95% of them, the models chose escalation. The AIs were trained on the most sophisticated strategic thinking humans have ever produced, and that literature, written to discipline irrational human beings, reads to an AI like a playbook. The model learned the logic, but it missed the lesson.

Today is National AI Literacy Day. Sean White shares why we need AI literacy and fluency to shape the future of AI. There’s a gap between AI literacy and AI fluency, and it's wider than most people realize. Literacy is knowing AI exists, knowing where it impacts life, and roughly how to use it for simple tasks. Fluency is knowing how to ask the right questions, how to push back on a bad output, how to use it as a thought partner rather than a search engine or ghostwriter. It's the difference between a calculator user and a mathematician. Better context and understanding leads to better results. There won’t be one AI to rule them all. That is obvious as AI adoption and literacy increase, the need for AI fluency also increases. And fluency isn’t about mastering one tool, it’s about knowing which tool to use, when, and why. We’re already seeing people use multiple AI tools for different purposes. We recently conducted a survey of 1,000 American consumers asking them about their tech habits. People who reported using AI said they have used 4 different AI tools. They reported using roughly 2 different AI technologies every day and 3 per week.

I'm proud to pledge $5,000,000 to Second Harvest of Silicon Valley's campaign to build a new home for the future—one that will help get nutritious food to more neighbors, faster. Through June 30, every dollar you give will be tripled. Join me in strengthening our community: https://lnkd.in/g8NMR6Ei 1 in 6 people in Silicon Valley rely on Second Harvest to stay fed. This new building will reduce costs and expand access for more of our neighbors. Silicon Valley has given us so much. It's up to all of us to give back.

I've been fascinated by the intersection of AI and music, particularly how easy it is becoming to create songs that explore different niche ideas and stories. I'm back to share another experiment... this time a song about April Fools'! April Fools' Day is a holiday with no founder, no committee, and no origin anyone can agree on. Nobody even knows where the apostrophe goes. And yet billions of people celebrate it every year, in every country on Earth, with no coordination whatsoever. I've spent my career thinking about networks — how people organize, build trust, and coordinate at scale. Usually that requires platforms, protocols, incentives. But April Fools' Day is a case study in something rarer: a global institution that emerged from no institution at all. No one planned it. People just showed up. That's what this song is about. "Nobody Knows" starts with the apostrophe confusion and the missing origin story — things that should be problems — and arrives somewhere I didn't expect: these aren't bugs, they're features. A holiday that can't explain itself is the most accurate portrait of how people actually work. Messy, undefined, no one in charge, functional anyway. That's awesome. Happy April Fools' Day — however you punctuate it!

Had a fantastic time joining Ray Dalio and Ben Casnocha on stage to talk frontier AI, the shifting world order, and the power of truthtelling in entrepreneurship. Thanks to Village Global for convening us!

I sang a song and produced a music video for Pi Day. Everything, including the music, video, and my voice, was created using AI.  Here’s how we made it happen, and what we learned: One of the things I find most exciting about this moment in AI is that it collapses the distance between an idea and a creative artifact. You can have a strange, specific idea—“what if we made a Pi Day music video?”—and instead of letting it die as a joke with friends (or in your head), you can actually make the thing. Not every use of AI has to be a coding agent sprint on an application that instantly generates millions in revenue. Some of the most important breakthroughs happen when people play—when they pick up the tools and build something just because it sounds fun. This project was very much in that spirit: an experiment that turned into a real test of a new kind of creative stack (shared below). The result is weird, fun, and very much an experiment. Exactly what it should be. I learned a lot from this process. More than anything, this reinforced something I keep coming back to: using AI to build in an unfamiliar domain is one of the best ways to learn both the craft itself and the capabilities of the tools. I’m not a songwriter. I’m not an animator. But working through this process taught me more about music production, animation pipelines, and creative AI workflows than any demo or keynote could. Today, it’s a music video about a mathematical constant. Tomorrow, this capability will power new businesses, new art forms, and new ways for people to share their ideas with the world. Let me know what you think—and happy Pi Day. — Here are the tools we used and how the pieces fit together: - Suno for music generation. - Kits.AI to transform the AI vocals into my voice. - Anthropic's Claude Cowork for planning, songwriting, and lyrics. - Logic Pro for audio editing. - Final Cut Pro for video editing. - Runway and Google's Nano Banana for image generation. - HeyGen IV for video generation of "Animated Reid". And for the lyrical animations, we used fal.ai’s Whisper model to generate timestamped transcriptions of the music track. Then we passed the transcripts to GPT-5.2 to generate video prompts. From there, we used an image-to-video model to create each snippet. After each one was generated, we extracted the last frame as an image and passed it into the video model with the next prompt. Then repeated that process across the full sequence.

CryptoPunks is what happens when you nail viral distribution. The numbers tell the story: roughly 1.41M ETH in total trading volume since launch—about $3.91B changing hands—with individual Punks now in prestigious collections from MoMa to LACMA to the Centre Pompidou. This week on Possible: we explore the world of Punks, and how their creators, Matt and John, stumbled upon viral growth and learned how to nurture it.

People say SaaS is dead, slain at the hand (neural network?) of Artificial Intelligence. Clearly, though, the $400b+ market hasn’t disappeared, but it is changing. The inference most people are drawing is wrong, and it’s worth being precise about exactly what’s in store for SaaS:

Notion’s Founder, Ivan Zhao, says you shouldn’t try to save costs or predict ROI when you’re implementing AI. In fact, it’s the easiest way to lose the race. A lot of companies are trying to “control spend” by rationing tokens and putting guardrails around how much experimentation engineers can do. But that creates artificial ceilings on what’s possible, right when the whole point of AI is to explore and discover new product surface area faster. Ivan’s argument is that AI flips the constraint inside a tech company. The bottleneck isn’t “can we build it?” anymore. It’s “what should we build?” Judgment and taste become scarce resources. If you artificially constrain what you’re able to try by limiting spend, you’re volunteering to fall behind.

AMD CEO Lisa Su has one piece of advice for all tech companies: Decide what you’re the best at. Some companies believe that with enough resources, talent, and ambition, they can be world-class across dozens of dimensions simultaneously. The most enduring companies, especially at early stages, are defined not by the breadth of what they attempt, but by how they clearly beat out the competition in their vertical. The companies who win are the ones who identify their single most important mission, and orient their entire organization –– measurement, hiring, planning, investments –– around achieving that objective. That requires making big long-term bets and having the conviction to stick with them even when the tech fads have yet to catch up. When everything is a priority, nothing is. That's not to say that a company can't be great at many things, but the question is whether you've earned the right to expand by first going impossibly deep on the one thing that makes your company and you irreplaceable.

Matt Hall and John Watkinson stumbled upon the most valuable real estate on the internet: the profile picture. It's one of the reasons they believe Cryptopunks became one of the largest digital art projects in human history: with roughly 1.41M ETH in total trading volume since launch—about $3.91B changing hands—with individual Punks now in prestigious collections from MoMa to LACMA to the Centre Pompidou. The underrated lesson here is about leverage. The moment your users start telling the story “as themselves,” it stops being a project you promote and starts benefiting from strong network effects. If you don’t have a budget (or you’re not naturally great at marketing), you can still win by making something that makes other people want to carry the message. Every new user becomes a channel, and every profile picture becomes a billboard.

I asked Notion’s Ivan Zhao a question: Notion is one of the few scaled software companies that’s genuinely making the transition from SaaS-native to AI-native, and doing it in public. How? Ivan’s answer was blunt: it’s very difficult, and almost no one has done it well. His edge, he argues, is that Notion already has practice reinventing itself. More on how Ivan is transforming Notion for the AI age:

Honored to be interviewed alongside visionary leaders who are shaping the AI future. It is my hope that this documentary contributes a small piece to making sense of where humanity and technology are today, and the future world AI is shaping.

Don’t put yourself in a box… unless it’s a Proto Hologram Box! We wanted to keep experimenting at the cutting edge of AI tech, so we turned ReidAI into a hologram. What stood out: it feels more human at first glance — the presence, the dimensionality, the illusion of being “there.” That means more people walk up to ReidAI and actually engage with it when it’s in the Proto Box—not because they think it’s me inside (I hope), but because the experience feels more intuitive. Bridging these interaction gaps with AI models and digital twins is an enormous opportunity and something that we need to think deeply about to feel the full benefit of AI in our lives. Also: Thanks to Village Global for hosting ReidAI and me!