Welcome back, Superhuman. The team behind a popular financial newsletter has been on a secret mission for three months: prompting Claude to find a new cure for scorpion venom. Apparently, they took this meme a bit too literally. Meanwhile, a new report offers more insight into Anthropic’s finances.
Today: Perplexity Computer comes to Mac, how to get an instant morning briefing from your iPhone Action Button, and get the latest prompts and trending social posts.
TODAY IN AI
Learn how Perplexity Computer lets you run AI on your Mac. Source: Perplexity.
1. Anthropic’s S-1 reveals an $80B loss: According to IPO documents reviewed by the WSJ, Anthropic posted half-year revenue of $16B, a net loss of $80B largely tied to accounting changes, and could spend up to $750B on compute over the coming years. Anthropic’s public listing is expected to land mid-November, which will offer better insight into the economics behind top labs. Before Anthropic’s listing invites comparisons, OpenAI is reportedly rushing to close another $30B funding round.
2. Perplexity joins Meta by bringing its models to Apple Macs: Perplexity Computer can now run locally on a Mac, giving users the option to use on-device AI when working with private files (36GB of RAM required). The tool also offers a hybrid approach: local models interact with computer files, while cloud models handle reasoning. The update signals a growing trend: Meta announced that its new Muse Spark Open is also optimized for Apple’s Mac Studio.
3. ElevenLabs and Krea roll out voice and image models: ElevenLabs released Eleven Realtime, a new class of speech-to-speech models designed for building natural-sounding voice agents like digital assistants, tutors, or characters. Separately, Krea released a new image model built for editing images at scale. It was first announced last week, but just landed at #4 on Arena’s Multi-Image Edit leaderboard, trailing only OpenAI.
PRESENTED BY WORKOS
Your app needs authentication. That means installing an SDK, wiring up callback routes, configuring middleware, and getting the environment right.
Run npx workos@latest install. The WorkOS AI installer detects your supported framework and adapts AuthKit to your existing project. It installs the SDK, creates auth routes, and sets up environment variables, with validation enabled by default.
Review the generated code, test your login flow, and get back to building.
TUTORIAL
How to get an instant morning briefing from your iPhone Action Button using Grok Bot
Click here to see a visual demo of how Grok Bot integrates with your iPhone Action Button. Photo: SpaceXAI
Open Grok Bot on your iPhone, update it to 1.16, and open a voice chat once so microphone permissions are already enabled.
Create a dedicated Bot for your morning briefing. For details on how to create and manage Bots, see here.
Connect the work apps it should read, such as your email, calendar, Slack, or other supported plugins.
Sample Prompt: “When I ask for my morning brief, review today’s calendar, recent important messages, outstanding follow-ups, and active priorities. Give me my top 3 priorities, meetings I need to prepare for, anything overdue or blocked, and one recommended first action. Keep the briefing under 2 mins and prioritize anything time-sensitive.”
On your iPhone, open Settings > Action Button, swipe to ‘Shortcut’, then tap ‘Choose a Shortcut’. Select the Grok Bot voice action exposed by the app.
Press and hold the Action Button each morning, then say something like “Give me my morning brief” to launch the Bot hands-free.
FROM THE FRONTIER
You can’t just prompt Claude to discover a new cure…can you? Citrini Research actually did the meme.
Click to read the full story behind Citrini’s mission to prompt a new drug using Claude.
Tech employees are smoking cigarettes again. They’re so confident that AI will cure cancer they figure, why not? Curing diseases has been one of AI’s biggest selling points for years. But how much of that is marketing hype vs. reality? It’s often hard to tell, so financial media firm Citrini Research tried to figure it out for themselves.
Citrini’s team of five people — none of them scientists — attempted to prompt their way to creating a new medicine using Anthropic’s Claude. They decided to pursue a scorpion antivenom, and made a shocking amount of progress very quickly:
In just three months, Claude created 211 potential antivenoms, 17 of which met Citrini’s criteria for success. Generating these results took about 367 messages, 10,000 H100 GPU-hours, access to an internal Anthropic model, and a $200,000 budget.
Claude handled essentially everything, from presenting initial ideas and assembling findings to emailing suppliers and writing its own prompts.
So did Claude find a cure? Not yet. Citrini’s candidates succeeded at sticking to their toxin, but weren’t tested to see if they can neutralize the venom. The firm also opted out of the next phases: animal studies, dosing work, and safety testing. Still, Citrini considered it win. Claude completed roughly one year of work in days, and took the team further than they could’ve ever gotten alone.
PRESENTED BY MONGODB
Your app will live (or die) by how relevant its outputs are.
Nix the constant refresh. With MongoDB, your apps automatically retrieve the latest data to deliver timely, up-to-date answers with less manual work.
IN THE KNOW
What’s trending on socials & headlines today
Meme of the day
📈 Claude’s Recs: As agents continue to drive internet traffic, ranking in chatbots’ answers is becoming a top priority. The CEO of an AI search company laid out 11 strategies for ranking in chat responses that got 8.5K bookmarks.
🔒 Locked Chats: OpenAI rolled out a long-overdue update: privacy. You can now lock any sensitive chats behind a 4-digit pin code. See how it works.
💨 Speedy PC: The internet’s been using AI to speed up laggy TVs. Now a similar trend is starting with computers. This viral thread outlining 7 prompts to speed up your PC has racked up over 25K bookmarks.
📉 Freelance Freefall: Are freelance web design gigs officially over thanks to AI? One Reddit post has sparked a discussion: the top comment agrees, but the second comment offers a refreshing counterpoint.
😆 TikTok Trend: People on TikTok are asking their favorite LLM to “list 5 things I always say.” The responses are all over the place — from gentle and friendly to surprisingly aggressive.
PRODUCTIVITY
🚀 G2*: Skip the slow surveys. Voice agents interview 100+ buyers a day and deliver insights 5x faster. See the demo.
💵 Composio Instant: Helps AI agents pay for their own tools.
☑ ToDo: Write a to-do task and an AI agent gets it done.
👧 Stella: An AI companion for emotional wellness and life planning.
💻 The Code: A 5-minute weekly brief to help programmers stay on the cutting edge.
PROMPT STATION
Design an effective product experiment
Prompt: Act as a senior product manager and experimentation strategist. Help me design a rigorous product experiment for [feature/change/hypothesis] that produces ...................IMAGE
Create a photo for your work accounts from one selfie with AI
Prompt: [upload a clear selfie] use the uploaded selfie and transform it into a clean, polished profile photo suitable for work accounts like Slack, Microsoft Teams, Gmail, ...............👉 Click here to open the full prompt in ChatGPT (Don’t forget to switch to Images!)
EXTRAS
Want more?
Learn how to use AI at work with our AI Academy
Check out our Top Prompts and Top 125 AI Tools
Level up with our newsletters for marketers, programmers, cybersecurity and IT
Grow customers & revenue: Join companies like Amazon, HubSpot, and Salesforce. Showcase your product to our 2 million+ readers and followers on socials. Get in touch.
What did you think of today's email?
Until next time — Zain, Theodore, & the Superhuman AI team.