Sign in Subscribe

OpenAI is on a collision course with Google to be the consumer productivity company. If I ran product at OpenAI, here’s what I would do. 

  • Thesis: OpenAI and Google are both fighting to be “the productivity company” for consumers (ChatGPT vs Google Search) and enterprises (OpenAI API vs Google Cloud). There’s a lot of talk about OpenAI competing with Meta. This is off base. I replace my search behavior with chat, not my social media behavior with chat. 
  • Actions: 
    • Split the organization into two business units: consumer and enterprise
    • Rebrand the company to Merlin. OpenAI and ChatGPT sound like scientific research institutes, not a consumer friendly tech company. Merlin or a different magician is a good direction. This also ends the “ClosedAI” meme. 

  • Thesis: The consumer AI moat is first-party user data. The more data my AI has about me, the more tailored the interactions, the more entrenched I’ll be in that ecosystem. Similar to the Apple and Google ecosystems. I don’t think consumers will switch between AI ecosystems for much longer as the AI models start to store user data and personalize. As ChatGPT learns more about me, I find myself using it a lot more, even if it’s slightly worse than Claude at some use cases. A simple example — I use ChatGPT a lot for recipes and it has learned my food preferences. This makes me way more loyal.
  • Actions: 
    • Prioritize collecting as much first-party user data as possible by building a full personal productivity suite. In priority order — Chat/Search, Email, Browser. Make all of these apps standalone, but gradually make the primary Merlin Chat the core interface. Eventually I shouldn’t have to go into email or a browser, my Merlin should do that work for me. 

  • Thesis: Chat is the new operating system and it will be two-tier. The tier one primary AI will delegate tasks to tier two specialized AIs. Similar to how an OS launches apps, or browsers launch websites. If you ask ChatGPT a math question, it will delegate it to the math-specialized AI. This delegation will become increasingly seamless and invisible to the user. The big winner will be the primary AI interface that launches sub AIs and builds an “App Store” of specialized AIs. Tier two includes agents as well. 
  • Actions: 
    • All the Merlin productivity apps need to be redesigned to be chat centric. Hire a small, elite team of brand product designers to lead this effort. Merlin needs to be a consumer tech company, not a research company. 
    • Merlin: Primary interface, delegates to other AI apps (AIpps pronounced apes). The core chat interface will hook into the other apps, so you never really need to leave chat. The core chat will need to be more visual to pull up data from the underlying apps.  
    • Merlin Mail: Prioritized queue, auto draft responses in your voice based on existing ChatGPT usage, eventually agentic auto-responses. Reduce 90% of time spent in email. 
    • Merlin Browser: Every webpage is personalized and chat interactive. Every NY Times article is explained to me personally and queryable to go deeper on specific topics. 
    • Merlin Messenger: I honestly don’t know what this would look like, and this gets more into “social” than “productivity” so Merlin might not be able to pull it off. Google never really did. But I would try, with the aspirations of making this like the Chinese super apps like WeChat.
    • Invest resources into the Merlin App Store to help companies build AI-native versions of their current software. Make chat the primary interface to the Internet, replacing mobile apps and the browser. 
    • Build a personal computing device to replace mobile phones. My bet is voice will be the core interface, so I’d do earphones + a watch for a small screen. The heuristic for a successful AIpp should be “can I use this with only my voice while driving my car? Can I read and respond to my email using this AIpp?

  • Thesis: Superintelligence happens and all of this could be completely wrong because my feeble mortal brain can’t even conceive of what a superintelligent future looks like.
    • Action: Proactive ritual sacrifices to our impending robot overlord
    • The superintelligence end state still requires transitionary bridges to get there, so the two-tier AI model probably is still valid. It might just not last very long. 

  • OpenAI product strategy summary:
    • Split OpenAI into two business units - consumer and enterprise. 
    • Rebrand to Merlin, become very consumer design oriented, escape the bad PR. 
    • Build the AI personal productivity stack: Merlin (combine the current ChatGPT and GPTSearch), Merlin Mail, Merlin Browser, Merlin Messenger. Eventually Calendar, Docs, Drive, etc too. 
    • Invest heavily in the Merlin App Store and onboard companies to build AI apps
    • Build a personal computing device focused on earphones to replace the mobile phone. Voice will be the core UX.
    • Hope that you can build an AI-native personal productivity company faster than Google can rebuild their personal productivity stack.
    • Maybe none of this matters and superintelligence completely changes the way we interact with technology, other humans, and the world.

  • Thesis: Start-up investing implications, ordered from least promising to most:
    • Invest in infrastructure for AIpps — analytics, hosting, model training, databases, etc. What will it take for 1 million AIpps to bloom? IMO this is too early; the app layer needs to develop first. Plus the cloud providers (AWS, Azure, etc) are in the best position to win this.
    • Invest in AIpps. The weirder the better. Break my millennial brain. What does a chat-native game look like? Who is building the Zynga AIpp? What about music production, gambling, messaging, shopping, travel? I would invest in things that are so weird they are more art than business. Creative, smart people will experiment and figure out the business model. Find me some weird Gen Alpha people making weird stuff.
    • Invest in AI-enabled physical tech. 8VC is the role model. Those businesses are more defensible against Mag7 swooping in.
    • Quit VC and invest in scarcity (real estate, natural resources). Robots building homes will increase demand for land. Robot farming will increase demand for land. Autonomous vehicles, EVTOLs will increase demand for land further from cities. Henry George already proved that progress’s value accrues to technologists and to landowners. Land ownership is much easier. (But very messed up that landowners accumulate wealth by doing nothing). Land value tax, you’re our only hope. 

  • Thesis: Mag7 investing implications, ordered from most bearish to most bullish
    • Apple: Bearish. AirPods + Watch should be the ideal AI interface, I’m just nervous that Apple won’t get it together. If chat replaces mobile as the core consumer platform OS, Apple is in a bad spot. Their lack of AI progress and declining growth makes me concerned. 
    • Microsoft: Neutral. Their productivity stack could be replaced, but their OpenAI ownership protects them. 
    • Meta: Neutral. I don’t know how social will shake out in a chat-first OS world. 
    • Nvidia: Neutral. AI will of course drive demand for more chips, I just don’t know enough about this market to know if there will be more competitors and how big Nvidia’s current moat is. 
    • Google: Bullish. Good shot at being the consumer productivity company. 
    • OpenAI: Bullish. I believe OpenAI can build an AI native productivity company faster than Google can rebuild theirs, as long as they don’t internally implode first.
    • Amazon: Very bullish. Physical world “bits” companies will be at an advantage. AI improves their operations without much risk of a usurper. I don’t see a company becoming better at real world logistics than Amazon, and all the imminent vehicle, drone, and robotics autonomy will just make them more efficient. 
    • Tesla: Very, very bullish. Autonomous robots are arguably the largest market opportunity ever. Physical world is defensible. 

OpenAI product strategy and AI investment thesis