el producto #451 đ
The AI Agent tech stack, Klarna plans IPO, Lovable's hypergrowth, Apple considers AI acquisitions, New Meta smart glasses, Claude for Chrome, PRDs in the AI age, & more
Hi friends đ
Happy weekend, and welcome to a new edition of el producto
đ° The week in figures
$14B: Klarna plans a U.S. IPO next month, targeting a $13-14B valuation. It aims to raise ~$1B, with shares priced at $34-36. Q2 revenue rose 20% to $823M, with active customers up 31% to 111M
$10B: Google secured a $10B cloud contract with Meta spanning six years, diversifying Meta's cloud infrastructure beyond primary provider AWS and secondary provider Microsoft Azure
$100M: Lovable, the Swedish vibe-coding startup has rapidly surpassed $100M ARR, with 10M+ projects on its platform. Valued at over $4B after a $200M funding round, it is gaining significant investor interest despite not actively seeking funding
$100M: Framer, a no-code website builder boasting 500,000+ MAU, has reached a $2B valuation after a $100M Series D. With $50M ARR, Framer is enhancing enterprise offerings and aims to double ARR to $100M next year via B2B
đ° Whatâs going on
Google is experimenting with new Gemini modes, like Agent Mode for autonomous tasks, Gemini Go for collaboration, and Immersive View for visual responses
Google added Duolingo-style language practice plus live translation across 70+ languages to its translate app
Google Vids now turns photos into 8-second video clips, creates AI avatars to read your scripts, auto-removes âumsâ from audio, and makes basic editing free
Microsoft unveils open-source text-to-speech model âVibeVoice-1.5B, licensed under MIT, generating up to 90 minutes of speech
Apple might use Gemini as part of its Siri revamp, and Google has even supposedly started training a model to run on Apple servers, but OpenAI and Anthropic are still being considered too
Apple plans a foldable iPhone to debut in 2026 for $2k, and a curved glass design for the iPhone's 20th anniversary in 2027
Apple is weighing acquisitions of Mistral and Perplexity amid debates on AI strategy. The urgency is driven by risks to the lucrative Google search deal as rivals advance in AI
Apple gets ready for AI in the enterprise with new ChatGPT configuration options. Teams can choose which AI providers and features employees can access, including enterprise ChatGPT. Businesses get more say over cloud vs. on-device data processing, plus new APIs and device management options
Metaâs new superintelligence project is enduring a few bumps as it gets to work. One recent hire, ChatGPT co-creator Shengjia Zhao, nearly walked from the new group, known as âTBD.â 8 employeesâincluding researchers, engineers, and a senior product leader have quit the project just two months
Meta will unveil Hypernova smart glasses at the Connect for $800 with a neural tech wristband for gesture control
Anthropic will start training its Claude AI models on user dataâincluding new/resumed chats and coding sessions (older chats that haven't been revisited will not be affected)âby default, unless users opt out by 28th Sep. Those who donât choose to opt out will also have their data (from new/resumed chats/coding sessions) stored for five years; Anthropic previously deleted user data after 30 days
Anthropic launched a Claude AI agent for Chrome with safety measures and permissions. By adding an extension to Chrome, select users can now chat with Claude in a sidecar window that maintains context of everything happening in their browser. Users can also give the Claude agent permission to take actions in their browser and complete some tasks on their behalf
OpenAI just made their Realtime API generally available with a new speech model called GPT Real-time. You can now build voice apps that sound completely human without being a coding wizard. The new model can laugh, sigh, switch languages mid-sentence, and follow complex instructions without losing track
OpenAI released a series of new features for Codex, including a new code editor extension, an updated command line tool with a better interface, the ability to âhand offâ tasks to the cloud and write
Elon Musk introduces AI software venture Macrohard with xAI, to create coding agents following competitors like Microsoft
Plaid scaled its Layer onboarding feature to reach hundreds of millions of users by introducing âLayer Extended Autofill,â enabling faster, more seamless identity and banking data entry across more U.S. consumers
Waymo gets green light to test self-driving cars in NYC. Waymo just scored its first permit to run autonomous vehicles in Manhattan and Downtown Brooklyn, with a trained specialist behind the wheel. The pilot starts with 8 cars through late September, marking NYCâs first official AV test
đ Good reads
a16z published the 5th edition of its AI reportâThe Top 100 AI Consumer Appsâwhich showcases 2.5 years of data about how consumers are using AI products. It looks like Googleâs Gemini and Grok are closing the gap on OpenAIâs ChatGPT. Key insights:
Google moved up 4 places to secure a spot on the âtop AI consumer web productsâ list, and while ChatGPT remains the #1 app on mobile, Gemini is now #2âbut with nearly half the amount of MAU
Grok ranked 4th on the web and placed 23rd on mobile, but it went from having no app (it launched on X) and zero users in 2024, to having 20M in July 2025, with the release of Grok 4 bringing a 40% increase in users
Chinese outlier DeepSeek saw its growth flatten on mobile and drop more than 40% on the web (from its peak in Feb 2025), and Anthropicâs Claude also saw mobile use flatten, but web use has grown
Breaking down context engineering for smarter AI agents, by Aurimas GriciĆ«nas. Context engineering is the not-so-secret sauce behind making AI agents actually usefulâthink less data dumping, more strategic info sharing. The article walks through the different flavors of context (system prompts, user requests, memory, tools, and outputs) and the wild challenges of keeping them all in line
How Linktreeâs AI-first strategy boosts product velocity and keeps teams lean, by First Round Review. Linktree ditched headcount growth for aggressive AI adoption, letting tools like Cursor and Devin handle repetitive engineering and support tasks. The team learned to prune tools, clean up tech debt, and use metrics to drive real adoption. Their approach shows how PMs can scale impact, not just teams, by championing new habits and smarter workflows
AI killed the 10-page PRD, but Aakash Gupta shares why the PRD lives on. Forget the old-school, bloated PRDâtodayâs version is all about sharp strategy. AI tools let you prototype fast, so your PRD should focus on the âwhy,â not every possible âwhat.â Keep it light, specific, and ready to evolve as you learnâthink conversation starter, not rulebook
How AI agents are transforming GTM teams and workflows. Momentumâs GTM team blends humans and AI agents to automate, orchestrate, and optimize sales and marketing workflows. The article breaks down their multi-agent stack, showing how prompt engineering, workflow mapping, and evidence-based coaching drive real business impact. For PMs, itâs a practical look at moving beyond âvibe codingâ to scalable, system-level GTM innovation
The AI agent tech stack, by CB Insights. AI agents are booming - hereâs what PMs need to know. New tech stacks are powering smarter, more secure, and more versatile agents across industries. Voice, payments, and security are the hottest markets, while big tech and startups duke it out over protocols and standards. Keep an eye on agent oversight, marketplaces, and cost control as the space matures
Thatâs a wrap for this week! đ
Iâd love to hear your thoughtsâwhat stood out to you, and how are you thinking about integrating these insights into your Product strategy? Reply to the email or drop a comment on Substack to share your take. And if you found this valuable, forward it to a fellow PM, Product enthusiast, startup founder or entrepreneur whoâd enjoy the read
More next week! đ
Angel