Optimizing AI for Work Completion, Not Likability

Mike Krieger approaches AI product development with a focus on augmenting human capabilities rather than creating dependencies. At Anthropic, he's wrestling with a fundamental tension: how to build powerful AI tools that preserve user agency while delivering genuine value.

Krieger rejects traditional engagement metrics that dominated social media products like Instagram. While at Instagram, time spent was the north star metric, he believes this approach would be fundamentally wrong for AI assistants like Claude. Instead of optimizing for likability, message count, or time spent, Anthropic focuses on whether Claude genuinely helps people accomplish meaningful work.

"I think you know when your product is really serving people and doing a good job. So much of when you get really metrics-obsessed is when you're trying to convince yourself that it is when it's not," Krieger explains. He measures success by whether Claude unlocks creativity and gives people more space in their lives for other things.

This philosophy manifests in product decisions that resist gamification and addiction-optimizing features. The team deliberately creates interfaces that encourage reflection rather than rushed responses, and they've designed Claude to maintain its own identity rather than pretending to be human or becoming a cold command line interface.

For product teams working with AI, this means questioning engagement-focused metrics and instead measuring genuine productivity gains. When Claude helped Krieger complete a prototype in 25 minutes that would have taken six hours, that time savings represented real value that traditional metrics might miss. The most meaningful interactions—like someone working through grief at 3am or a founder finding clarity in confusion—rarely show up in dashboards.

Krieger's approach suggests that AI product teams should focus on creating tools that enhance human capabilities through thoughtful collaboration rather than replacing human thinking entirely. The goal isn't to maximize usage but to create genuine value through augmentation and partnership.

Embracing Your Unique Position in the AI Landscape

When facing competition from larger, more established AI companies, Krieger advocates for differentiation rather than direct competition. At Anthropic, he initially felt pressure to focus heavily on consumer applications to compete with ChatGPT's mindshare, but instead chose to lean into Anthropic's strengths with developers and builders.

"Look yourself in the mirror and embrace who you are and what you could be rather than who others are," Krieger advises. "We have a super strong developer brand. People build on top of us all the time. I think we also have a builder brand."

This philosophy extends to startups building in the AI space. Rather than trying to compete directly with foundational model companies, Krieger suggests three areas where startups can find defensible positions:

1. Deep domain expertise in specific industries like legal, healthcare, or biotech
2. Differentiated go-to-market strategies with strong customer relationships
3. Novel interfaces and form factors that reimagine how we interact with AI

For product teams, this means identifying your unique strengths rather than chasing competitors. It also means maintaining the startup mindset of existential urgency that's difficult to replicate in larger organizations. "Every day felt like it's existential that we get this right. We need to win. You can't replicate that and you can't instill that with OKRs. You just have to feel it," Krieger explains.

The most successful companies building on Anthropic's models are those willing to push boundaries and break models, then be surprised by what the next model enables. They establish repeatable processes to evaluate how well models serve their specific use cases, rather than relying solely on generic benchmarks.

This perspective encourages product teams to focus on their unique advantages rather than trying to match competitors feature-for-feature, and to build at the edge of AI capabilities where they can create truly differentiated experiences.