Welcome back, Superhuman. All systems go. SpaceXAI just launched ahead on the leaderboards with a new Grok model that rivals top competitors at a cheaper price. And Elon’s not done — he says an even better version will be ready in just a few weeks, though you should probably take his predictions with a grain of salt.
Today: Claude gets better internet skills, how to build an AI agent engineering team (without hiring anyone), and get the latest prompts and trending social posts.
You no longer need an X account to see X posts! Thanks to reader feedback, we will continue hosting all X (formerly Twitter) links on our website moving forward. If there are any other features you’d like to see, please let us know in the feedback section at the bottom.
TODAY IN AI
Learn about Claude’s new internet skills. Source: Anthropic.
1. SpaceXAI releases a frontier-level model one day after launching Grok Bot: Grok 4.6 offers frontier intelligence for knowledge work and coding tasks, matching OpenAI’s GPT-5.6 Sol and only trailing Claude Opus 5 and Fable 5 on the Artificial Analysis Intelligence Index. Unlike many frontier models, Grok 4.6 is highly cost-effective with a cost per task that’s unchanged from Grok 4.5 and comparable to open weight Chinese models. See how Grok 4.6 stacks up.
2. Anthropic’s Claude gets even better at using the internet: Claude Cowork — Anthropic’s desktop agent — can now access the web via Claude in Chrome, its side panel assistant for Chrome. This allows you to kick off tasks in Claude and continue them in a web browser side panel, where the AI can click, type, and scan pages for information. The update is rolling out now on paid plans. See how it works or learn to use Claude in Chrome safely.
3. Google unveils Pixel 11 lineup and a wave of device upgrades: At Made by Google 2026, the tech leader introduced its Pixel 11 series, powered by a new Tensor G6 chip. Gemini can now handle multi-step tasks across more than 40 apps, while a Live Translate feature delivers real-time translation of videos and audio. Google also rolled out updates to its Pixel Watch and Pixel Buds, along with a competitor to Apple’s AirTag. Read the full rundown here.
Yesterday’s most-clicked story: SpaceXAI rolled out Grok Bot, an “always on” desktop agent that can complete tasks for you by connecting to your most-used apps, tools, or websites.
PRESENTED BY WISPR
Wispr Flow turns your voice into clean, sendable text in any app, not rough transcription but text you actually send.
Reid Hoffman sends 89% of his messages with zero edits using Flow. Millions use it daily, including teams at OpenAI, Vercel, and Clay.
Why AI power users love Flow:
Works inside Cursor, Claude, ChatGPT, and other tools. Speak prompts 4x faster than typing.
System-wide on Mac, Windows, iPhone, and Android. One tap to start.
Understands code and context. Dictate into your IDE or send prompts as-is.
No cost on Android during launch. Try Wispr Flow
FROM THE FRONTIER
Claude’s latest update has sparked an unexpected controversy
Made with Midjourney
What’s going on? Anthropic plans to start digitally watermarking all Claude-generated content, a decision it made to stay compliant with a new transparency requirement from the European Union. The goal is to make it easy to identify AI-generated content — but not everyone is on board.
The point of contention: This update will slap everything Claude touches with a digital “created by Claude” label.
If you tell Claude to write a blog post, that blog post gets a watermark.
If you write a blog post 100% by hand but have Claude spellcheck it, that entire post still gets a watermark.
Many feel this will lead to lots of human-created work getting attributed entirely to an LLM, especially because most employees are strongly encouraged to use AI. Others have questioned how this will impact enterprise use cases, where AI is one small part of a deeply collaborative process.
Where do we go from here? OpenAI and Google have also committed to the EU’s transparency rules, so expect similar policies from each, though both already use watermarks for images and audio. It’s hard to believe, but with how quickly AI content is scaling, labeling human work might soon be the easier solution.
Kevin Chan is one person. He is replacing 20+ SaaS tools for a 50-person business.
His method: an agent inbox per workflow instead of a subscription per workflow
Install the Nylas CLI and authenticate with your API key.
Give each workflow its own inbox: nylas agent account create [email protected]
Repeat for invoices@, offboarding@, support@.
Limit what each one can act on: nylas agent rule create --name "Carriers" --condition from.domain,is,spam.io --action block
Hand the set to your coding agent: nylas mcp install --assistant claude-code
Or do it all with one prompt: "Create Nylas agent accounts for renewals@ and invoices@, block senders outside our carrier domains, then list the roster."
See how Kevin built it
PRESENTED BY AIRTABLE
Build fully custom sites, web apps, and interfaces on top of your Airtable data. Type a sentence and watch it take shape, styled and ready to share. When your data changes, it keeps up.
IN THE KNOW
What’s trending on socials & headlines today
Meme of the day
🤘 Breaking Barriers: DeepMind just gave deaf people a voice. Its breakthrough new model translates American Sign Language into written text by watching and interpreting visual hand signals. Learn how it works here.
🥹 AI’s Little Guys: AI mascots are getting noticeably cuter. That’s likely not a coincidence, but rather a clever marketing strategy to make AI feel more approachable and friendly. Watch the viral Grok Bot animation that sparked the conversation (1M views).
🤦 AI Overload: A cafe’s activity sheet for kids has gone viral as a great example of how not to use AI. It looks fine at first glance — but becomes more nonsensical the longer you look at it (19K likes).
👀 Enron Comparisons: The Big Short investor Michael Burry is sounding the alarm over Nvidia’s $500B funding deal, saying it has “echos of Enron.” This article breaks down his reasoning.
⏳ Revisiting History: The companies spending billions on AI’s infrastructure will probably go bankrupt building it. That’s the argument behind a viral post that claims this same scenario has played out a surprising number of times throughout history.
PRODUCTIVITY
IMAGE
3D paper-craft travel illustration
Source: Naiknelofar788 on X
ChatGPT Image 2.0: Create a charming miniature 3D paper-craft travel illustration of [DESTINATION], designed like a delicate handmade travel postcard. Feature recognizable local landmarks and cultural elements as tiny handcrafted models, arranged on a narrow floating landscape strip. Include small details such as local transportation, trees, street lamps, water, birds, clouds, and subtle travel elements like an airplane with a dotted flight path.
Use a soft white textured paper background, gentle natural lighting, pastel colors, subtle shadows, and a whimsical handcrafted aesthetic. The landmarks should look like they are made from paper, clay, wood, and miniature craft materials, with realistic tiny textures and charming imperfections.
Keep the composition minimal, clean, airy, and centered, with plenty of negative space around the miniature scene. Add a small elegant handwritten-style title at the bottom reading “[DESTINATION]”.
Style: miniature diorama, handcrafted paper art, travel postcard, whimsical 3D illustration, soft pastel palette, tactile textures, macro photography feel, delicate details, cozy artistic aesthetic, high resolution, vertical 9:16 composition.EXTRAS
Want more?
Check out our Top 125 AI Tools
Get better at prompting with our Top 1,000 Prompts
Learn how to use AI at work with our 50 Tutorials
Grow customers & revenue: Join companies like Amazon, HubSpot, and Salesforce. Showcase your product to our 4 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.
