el producto #414 🚀
Hard questions for leaders, Onboarding metrics, Strong network effects, Google's AI updates, Siri adds ChatGPT, Amazon Autos, Reddit's search Product, Usability testing tools & more
Hi friends 👋
Happy weekend and welcome to a new edition of el producto
🎰 The week in figures
$400B: Elon Musk's became the first person to surpass $400B net worth as SpaceX became the world's most valuable startup at $350B, and Tesla stock hit record highs on self-driving optimism. Special funds now offer SpaceX secondaries
$13.3B: Cyber Monday sales hit a record $13.3B, while Black Friday sales hit a record $10.8B. TikTok Shop also achieved a massive $100M+
$6B: xAI, Elon Musk’s SF-based AI startup officially closed its round at $6B at a $50B valuation. Its total raised is now $12B
1.7B: Twitch is still losing viewers with live viewership down to 1.7B from a pandemic high of 2.4B
$300M: One, the consumer fintech platform mostly owned by Walmart, raised a funding round over $300M at a $2.5B pre-money valuation
$200M: Zest AI raised a $200M growth investment to accelerate the adoption of its ML tools for credit underwriting
$135M: Public, the NY-based commission-free investment platform raised $105M Series D-2 and $30M debt led by Accel. Its total raised is now $400M+
€100M: Upvest secured €100M in a Series C funding round to enhance its investment API services for embedded finance platforms
$78M: Speak, an OpenAI-backed language-learning app, raised $78M in a Series C, doubling its valuation to $1B. The app uses generative AI to teach languages through audio conversations and speech drills, focusing on practical communication. With over 10M downloads, Speak aims to expand from teaching English into Spanish and French while maintaining its focus on efficacy over gamification
85%: AI agents can now predict your behaviour with 85% accuracy. A study by Deepmind and Stanford University conducted 2 hour interviews with 1,000 humans to understand their life experiences, beliefs and values. The 85% accuracy rates are much higher than the 70% associated with standard demographic modelling. The impact of this could be profound for use cases like conducting user / market research
📰 What’s going on
Google unleashes wave of AI updates: As part of a package of news items, Google dropped Gemini 2.0, which it claims will help push agentic AI apps ahead, a ‘Flash’ version of Gemini 2.0 that is better than Gemini 1.5 flagship and cheaper, an AI assistant for coding called ‘Jules,’ a new AI chip that is faster and consumes less power, and its new Android XR platform to eventually power Astra AI-smart glasses. Phew. Got all that? We’re still reading through all the releases, but Google’s AI just planted a new flag in the AI race to close out the year
In a Nature article, Google announced that its new ‘Willow’ quantum chip with 105 physical qubits could complete a task “in roughly 5 minutes” that would take “the world’s largest supercomputer an estimated 1025 years.” That’s massive. The change between Willow and other quantum chips is that it packs more physical qubits into a single package. As Google notes, “Willow can reduce errors exponentially as we scale up using more qubits,” adding that the development “cracks a key challenge in quantum error correction that the field has pursued for almost 30 years.” Implications for both quantum computing and existing software applications like encryption could prove incredible
Google Docs has rolled out a collection of 40 new templates, many of which should come in handy for Product teams (e.g. roadmaps)
iOS 18.2 adds ChatGPT to Siri, plus image and writing tool
The training data behind OpenAI’s new video AI Sora is raising eyebrows. Early users report it can generate clips inspired by Minecraft, Call of Duty, and even Super Mario Bros.—suggesting its training set may include Twitch streams and game footage. Some outputs even resemble popular streamers like Auronplay and Pokimane. Legal experts warn this could lead to copyright battles, with layers of IP at stake
Yelp is rolling out new AI aggregated reviews which let users filter based on sentiment. It’s called Review Insights and it uses LLMs to allow users to filter reviews based on things like quality of its food, service, ambiance, wait time, and drinks. Even when a review doesn’t explicitly mention one of the topics, the LLM understands the context in the review to identify and surface the relevant topic and sentiment. This type of AI-powered filtering could be useful for other use cases too
Reddit has been building a search Product based off its own data. The company’s Reddit Answers service will use a custom AI chatbot to answer user queries. Given that Reddit’s huge bounty of human questions, answers, and commentary is nearly without compare online today, conserving the value inside its own walls instead of selling them for pennies is a bold move but not one without risks. Google, after all, is a key source of Reddit traffic
Amazon is entering the car sales market with Amazon Autos, starting with Hyundai in 48 U.S. cities. Customers can search for new cars, secure financing, and schedule pickups online. The platform promises transparent pricing with no haggling, with plans to expand to more cities and automakers in 2025
In a surprising move, (even to the Cruise CEO) General Motors announced it would cease funding the robotaxi unit after $10B invested and years of R&D. GM is not alone in its retreat, Ford shut down its Argo AI self-driving venture in 2022, and Apple abandoned its car project earlier this year
📚 Good reads
25 hard questions every founder should ask themselves. First Round Review collected feedback by founders from all across the startup ecosystem and put together an insightful list of questions for anybody building Products (not just founders). This list will help you taking a pulse check on everything from Product strategy to co-founder relationships to their own decision making
The leaky bucket theory of Network Effects. Dan Hockenmayer explains how assessing a marketplace's ability to “hold water” in its bucket indicates its long-term viability. Narrowly defined markets may be easier to develop but are less defensible long-term, while successful marketplaces, like Amazon, are defined by their deep and diverse supply
Why Slack’s engineers break stuff on purpose. “Not only is it simply exciting to diverge from the usual job of fixing - it puts us in unique and realistic situations we would have never discovered if we had approached maintenance the traditional way”, says Slack Engineer Sean Madden
The only onboarding metric that matters is time to “Aha!.” Avi Siegel explains how different B2B and consumer products define their “Aha” moment, and why it’s essential to measure how long does it take for a new user to go from the very beginning of the product journey to the point where they realize yours is truly the product for them
Tools for conducting unmoderated usability testing. Unmoderated testing is a great way to get actionable insights from users quickly. In an unmoderated user test, there is no facilitator. Researchers set up questions and tasks in testing tools, and then participants complete those tasks asynchronously on their own. This collection of tools gives you some practical ways to do your own unmoderated testing. I love this comparison sheet. One to bookmark
5 different research interview question types. The questions that you ask in an interview can dramatically impact the insights that you glean from the interview. In this article, Rory Madden shares 5 different question types and the quality of the responses that you will get from each one
Product Management is broken; engineers can fix it - how Posthog redefined the PM and engineer relationship. An interesting yet slightly shallow post by Posthog CEO. While empowering engineers is certainly the way to go (I often say my job as a PM is to make myself redundant), the PM responsibilities at Posthog seem somewhat limited, and their approach works under the assumption that their engineers will rock at business thinking. Nice intentions, but make sure to adapt to your own context and team talent
That’s all for this week. As usual, feel free to reach out and share your thoughts by replying back to this email or commenting on Substack
Thanks for your support!
Angel