By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
EO Studio30-Minute Masterclass on Product Thinking | Instagram Co-Founder & Anthropic CPO, Mike Krieger
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
Great conversations run long. Their best ideas don't have to.
uListen is free on iPhone, iPad, Android, Chrome, and Claude.
See howCHAPTERS
Jump to publisher timestampsChapter summaries
Advice on Knowing When to Call It
Mike Krieger opens with advice for entrepreneurs deciding when to shut a company down, urging them to check in with their investors. He then introduces himself as Anthropic's Chief Product Officer and Instagram co-founder.
- •Entrepreneurs often feel they must keep going because investors expect it
- •Investors may say you've tried everything and it's time to call it
- •Closing a company is never a celebration but needn't be a tragedy
- •Krieger co-founded Instagram, started Artifact, and joined Anthropic under a year ago
A Brazilian Childhood and the Stanford Foundations
Krieger recounts how his dad's DOS computer sparked an early love of taking things apart, and how he only saw tech as a career after reaching California. At Stanford's Symbolic Systems program he took away lessons about solving real problems, prototyping and choosing the right partners.
- •Early curiosity sparked by a DOS computer at age four
- •No tech industry or role models in Brazil at the time
- •Symbolic Systems blended computer science, design, philosophy and psychology
- •Lessons: solve a problem for somebody, prototype early, find complementary partners
From Burbn to Instagram: Timing and Technology Waves
Krieger describes the 2009 landscape of weak phone cameras and early mobile apps, and how Burbn was stripped down to its photo-sharing core to become Instagram. He explains how to spot technologies ready for broader adoption, using the pace of iPhone improvement and Claude model leaps as examples.
- •In 2009 phone cameras were poor and few new social products existed
- •Burbn was a location check-in app that was good but not the destination
- •Stripping away distractions revealed photo sharing as the working core
- •Look for early adopter activity and a fast rate of change
- •Build for today while ready to catch the next big model leap
Joining Facebook and Building the Right Team
After the Instagram acquisition, Krieger talks about joining a larger company that tried to preserve startup culture through hackathons and experimentation. He stresses the value of a team of talented, low-ego generalists who care deeply about the product.
- •Instagram was 13 people with no built-out revenue before the acquisition
- •Facebook preserved startup culture with company-wide hackathons
- •Team members who are talented but low-ego are hard to find
- •Generalists and fluid roles, such as engineers working across backend and mobile, matter
- •Great products come from people close to the details and passionate about them
Artifact: Personalized News and Why It Stalled
Krieger explains the 2021 founding of Artifact, a bet on combining machine learning and product design for personal experiences, starting with news. After a two-year private beta and a year live, the product lacked energy because of poor destination sites and a weak experience before personalization kicked in.
- •Thesis: machine learning could make products feel personal, starting with news
- •Private beta lasted two years, too long, and exhausted the team
- •Linked mobile web sites were full of ads and poorly formatted
- •Personalization only became compelling after reading many articles, so users bounced
Lessons from Artifact Applied at Anthropic
Krieger says intelligence alone won't drive adoption; products must solve problems from the start. At Anthropic this shapes efforts to lower the barrier for newcomers and gather user intent quickly through onboarding questions.
- •Users won't adopt a product just because the technology is intelligent
- •Claude must be useful for someone discovering LLMs for the first time
- •Show concrete things people can do today rather than in theory
- •Onboarding questions help personalize Claude quickly, a direct result of Artifact
How to Decide When to Shut a Company Down
Krieger describes how he and Kevin listed the ideas they would regret not trying, prioritized three, and committed to stopping if they didn't change the trajectory. He repeats his advice to check in with investors rather than staying on out of obligation.
- •List the ideas you'd feel silly not having tried before closing
- •Prioritize about three big ones and assess whether they shift the trajectory
- •Artifact wrapped up the experiments around the turn of 2023 to 2024
- •A concrete date or set of projects helps you know when to move on
- •Check in with investors, who may agree it's time
Building Products Alongside Evolving Models
Krieger describes the mix of zero-to-one and established work at Anthropic, including the Claude character team and the difficulty of assessing a model's vibes. Product and model are both moving targets, with models changing up to a week before launch.
- •Anthropic still had a lot of empty product space, such as mobile apps and Claude Code
- •The Claude character team focuses on Claude's philosophies and vibes
- •Vibes are hard to evaluate; hundreds of internal conversations happen daily
- •Model taste in features like Artifacts shifts from model to model
- •Both product and model are moving targets, making development more challenging
Setting Expectations and Gathering User Feedback
Krieger explains the importance of being upfront about model capabilities, limitations and risks, and of Claude recognizing its own limits. He also describes how thumbs up and thumbs down feedback with written reasons is aggregated to guide each new model, without training on user data.
- •Onboarding emphasizes what models are good at and where they err
- •Claude noting it isn't a doctor reflects awareness of limitations, not just liability
- •Written reasons behind thumbs up/down are the most valuable signal
- •Aggregated themes, such as verbosity or caving too quickly, shape the next model
- •Anthropic does not train on user data, including conversations
Industry Consolidation in AI Models and Apps
Krieger predicts consolidation on the model side to perhaps three to five companies because of compute costs, with Anthropic well-positioned through partnerships. He expects app-layer consolidation as well, as happened with mobile and social media.
- •Frontier models require ever-larger compute and resources
- •Likely three, four or five companies will build them
- •Anthropic has partnerships with Amazon and Google
- •Y Combinator startups are nearly all AI-related, and not all will work
- •App consolidation may become visible in roughly six months
The Future of Claude: Memory, Proactivity and Agency
Krieger lays out where models need to go: understanding people beyond a single conversation, knowing when to be proactive, and acting independently as long-running background agents. He uses Claude in Slack as an example of the proactivity challenge.
- •Models need memory and contextual understanding for longer-horizon empathy
- •Models must learn when to chime in and when to sit back
- •Claude in Slack either participates too little or says too much
- •Agency means taking direction and working in the background over longer time horizons
AI as a Coach and Evolving Relationships with Models
Krieger envisions models giving people insights about themselves, acting as an ongoing coach similar to the coach he has worked with for years. He describes using Claude to poke holes in his strategy documents and wonders whether models will feel like friends or coworkers.
- •He has worked with a coach for years and wants that for everybody
- •Current coaching with Claude is point-in-time; the aim is ongoing
- •He uses Claude to ask what he's missing in strategy docs
- •Speculation on whether long-horizon models become friends or coworkers
Advice for the Young and Searching, and Staying Curious
Krieger tells people in their early 20s to trust winding paths and invest in relationships that recur throughout a career. He shares his hope that Anthropic is his last job, his habit of wanting each year to feel different, and his view of entrepreneurship as imagining how the world could be better.
- •Early 20s offers a unique vantage point between teenage trends and the ability to build
- •Don't worry that every step must make logical sense
- •Relationships built in a career recur over and over
- •Check in yearly on whether you're still learning
- •Entrepreneurship is getting curious about what could be different, and AI will empower people to ask and answer that
