el producto #345 🚀
Making Freemium Products, Meta's AI-code-assistant, BNPL vs SNPL (Save Now, Pay Later), The Next Job framework, Unlocking service marketplaces, Apple Watch X, & more
Hey team,
Happy weekend and welcome to a new edition of el producto!
🎰 The week in figures
$100M: BitGo, a Palo Alto-based crypto key protection startup, raised $100M at a $1.75B valuation from undisclosed investors
$100M: Anthropic is raising $100M from South Korea’s SK Telecom and will partner with it to build a global telecom-focused LLM
$95M: ClassWallet, a digital wallet-based purchasing and reimbursement platform for public funds, raised $95M
+25%: The average subscription cost of ad-free streaming services is up ~25% in about a year as streaming giants push customers towards ad-supported plans
-25%: Trading in Adyen shares briefly halted after 25% price loss; the Amsterdam-based FinTech unicorn published its financial results for H1 2023, and although it reported revenue of $804.3M, up 21% from 2022, the results came in lower than analysts had estimated
13%: Prior to the November launch of ChatGPT, only 7.7% of LinkedIn users claimed AI skills, but within seven months, that number almost doubled to 13%
📰 What’s going on
Meta to release free AI software that generates code, competing with similar offerings from OpenAI and Tabnine; "Code Llama" is open-source and could launch as early as next week; the system can offer automatic code suggestions to developers as they type; companies can use it to create their own AI assistants offering code suggestions
Apple is developing an updated "Apple Watch X" featuring a microLED screen and blood pressure monitoring; the release of the watch is expected in late 2024 or 2025, coinciding with the Apple Watch's 10th anniversary
Former Google CEO Eric Schmidt is building a nonprofit organization, modeled after OpenAI, to take on science/biotech challenges with AI solutions
Google’s combined AI team is working on performing 21 different types of personal and professional tasks with its generative AI models, including life advice, ideas, planning instructions, and tutoring tips
Amazon is expanding its buy-now-pay-later (BNPL) service to other retailers that use the Amazon Pay checkout tool; the BNPL service will become available at online stores, including Lenovo and others
X (fka Twitter) will no longer let advertisers promote their accounts on user’s timelines; promoted accounts generate $100M+ revenue annually for the company
Substack will let users follow writers without subscribing to their newsletters in a further expansion into social networking
Former Intuit SVP Alex Chriss, who played a significant role in Intuit’s $12B Mailchimp acquisition, has been named CEO of PayPal, replacing longtime CEO Dan Schulman
📚 Good reads
Unlocking service marketplaces. Why service marketplaces aren’t scaling and how they might get unstuck. Great read by Dan Hockenmaier on the evolution the next holy grail of consumer marketplaces, that promise to unlock new major opportunities in the space. Dan does a great job in identifying that AI can have an impact in getting services marketplaces to deliver full transactions instead of lead gen only
Save Now, Buy Later: what is SNBL, and why is it causing such a stir across consumers, retailers, and FinTechs alike?
A guide to making Freemium Products, by Elena Verna. “Freemium is a business strategy, while trials are conversion rate optimization tactic - you shouldn’t compare the two. You should, however, use them both. You can capture the best of both freemium and trial worlds in a freemium with a reverse trial“
The Next Next Job, a framework for making big career decisions, by Andrew Chen. “What do you want to be your next next job? And why can’t you get it right now?”
Productboard’s path to Product-Market Fit, — building a $1.7B unicorn company with Lean Startup principles, by First Round Review & Productboard’s founder and CEO Hubert Palan
🔮 Emerging trends
How top AI models perform
Arthur, a platform for monitoring machine learning models, has released new research gauging how top large language models perform in areas like mathematics, so-called "hedging," and their knowledge of U.S. presidents:
OpenAI's GPT-4 performed best on questions involving combinatorial (counting) mathematics and probability, followed by Anthropic's Claude 2. Cohere's model performed the worst in math, with zero correct answers and 18 hallucinations, which occur when models generate inaccurate or nonsensical information
With U.S. presidents, Claude 2 got the most answers right, with GPT-4 in second place. Overall, the models produced over 120 hallucinations but didn't avoid any questions
the researchers examined how often the AI models respond with "hedging" answers, where their responses contain warning phrases to avoid generating offensive or false content (like "unfortunately, I cannot answer that question"). GPT-4 showed a 50% relative jump in hedging answers over GPT-3.5, which backs up some user experiences about GPT-4 being more "frustrating" to use than its predecessor, according to Arthur. Notably, all of Cohere's responses lacked any hedging language
The takeaway: GPT-4 outperformed all models tested, with fewer hallucinations than GPT-3.5, particularly in math questions where it reduced hallucinations by up to 50%. In contrast, Meta's Llama 2 experienced a higher overall rate of hallucinations compared to both GPT-4 and Anthropic's Claude 2
That’s all for today! Let me know what you think by replying back to this email or commenting on Substack
Angel