Good morning, {{ first_name | AI enthusiasts }}. Anthropic's first model in its '5' tier is here… But it won’t be holding the spotlight for long.
The newly released Sonnet 5 is an upgrade over its predecessor, but not at the levels we’ve become accustomed to — and it arrives at a bit of an awkward moment with the far more powerful and celebrated Fable and Mythos models now officially set to return.
Editor’s Note: Anthropic announced Tuesday night that the Department of Commerce has lifted its export controls on Fable 5 and Mythos 5, with the models set to come back online Wednesday. We'll have deeper coverage on this in the next edition.
In today’s AI rundown:
Anthropic's Sonnet 5 arrives in Fable's shadow
Google’s new Nano Banana Lite, Omni Flash models
Create winning ads with Claude in one command
Anthropic debuts new science hub for Claude
4 new AI tools, community workflows, and more
LATEST DEVELOPMENTS
ANTHROPIC
Image source: Anthropic
The Rundown: Anthropic just launched Sonnet 5, the latest upgrade in its mid-tier model class that it calls “the most agentic Sonnet model yet” — coming just before the U.S. Dept. of Commerce lifting export controls on Mythos 5 and Fable 5 after 18 days.
The details:
Sonnet 5 shows major jumps in agentic coding and reasoning over its predecessor, with knowledge work capabilities that surpass even Opus 4.8.
The new Sonnet can operate a browser or terminal and carry longer jobs, bringing more Opus-style agent behavior into Anthropic's cheaper tier.
The model’s cybersecurity benchmarks come in worse than Sonnet 4.6, with Anthropic saying it “did not deliberately train” 5 on cybersecurity tasks.
Sonnet 5 is available across plans, with API rates priced at $2/$10 per M input/output tokens until Aug. 31 and $3/$15 after.
Why it matters: It’s an upgrade over 4.6, but the backdrop of users longing for Fable leaves this feeling like an underwhelming start to the ‘5’ class. Many releases are accused of ‘benchmaxxing’, but Sonnet feels like the opposite for cybersecurity — an awkward, direct result of the Fable and Mythos entanglement the company has been in.
TOGETHER WITH MERCURY
The Rundown: You've seen what AI can do when it's built into the right system, but your finances probably aren't there yet. Mercury Command is AI built directly into your banking* that turns plain language into executed work across your account — with your approval required at every step.
Ask Command to:
Send invoices and categorize transactions
Summarize what needs your attention before month-end
Send a payment (so you don’t have to)
Image source: Google
The Rundown: Google just introduced two new media models into its API for developers, with a lower-cost Nano Banana 2 Lite for fast, bulk jobs and a powerful Gemini Omni Flash with strong video generation and editing capabilities.
The details:
Lite can generate an image in just four seconds at $.034 per image, aimed at cost-effective, bulk workflows at quality below the frontier.
Gemini Omni Flash also rolled out to developers, a model that generates and edits 10-second video clips at $.10/sec that tops text-to-video leaderboards.
Omni Flash trails only Seedance 2.0 in video editing, with Google injecting Gemini’s multimodal reasoning into the video model for real-world knowledge.
The pitch is to link the two models, with users able to make an image with Lite, provide it to Omni, and then animate it into a video clip in a single workflow.
Why it matters: Some may be disappointed not to see a massive new jump, but Google is putting capabilities into users’ hands at speed and prices that make creative workflows stronger than ever. With Google’s sprawling ecosystem of apps to power, those qualities are vital to pushing everything forward at scale.
AI TRAINING
The Rundown: In this guide, you will learn how to use Claude to scan your website, extract your brand DNA, and generate ad creatives with the Goose Ads skill.
Step-by-step:
Open a new Terminal window, paste “npx gooseworks install --all”, hit Return, and follow login or approval steps. This installs the Goose Ads skill for Claude
Type claude and press Return to start a new Claude Code thread. Then run “/goose-ads create ads for my brand [brand name] [link to your website]”
Claude will create a new project in the Goose Ads site and send you a link. If you click it, you can see all the brand info the skill is extracting from your site
Prompt: “Pick 3 templates from Goose Ads, estimate credit usage, and wait for approval.” Once approved, Claude will run generation and deliver the final ads
Pro tip: For the most cost-effective setup, use GooseWorks to generate draft ads set to "low" quality, then refine the best concepts with ChatGPT image or Nano Banana.
PRESENTED BY SCRIBE
The Rundown: Scribe Optimize automatically maps how work actually happens across your org — so before you build an AI strategy, you're working from reality, not assumptions.
With Scribe Optimize, you can:
Map real workflows — no surveys or consultants
Surface inefficiencies leadership can't currently see
Build your AI strategy on ground truth, not gut feel
ANTHROPIC
Image source: Anthropic
The Rundown: Anthropic released Claude Science, an AI workspace for scientists that brings paper review, databases, figure-making, and computing jobs in a unified hub — with the company also starting its own preclinical drug discovery program.
The details:
The platform is designed to link and trace every step of the research process, keeping an auditable track for better understanding work and fact-checking.
The app connects 60+ scientific sources and tools for genetics, proteins, chemistry, and cell data, cutting down on manual database hopping.
Claude Science is rolling out across paid plans on both MacOS and Linux, with sensitive datasets able to stay on a lab's own machines instead of the cloud.
Anthropic is also starting its own drug discovery effort targeting “neglected” diseases traditionally skipped by pharmaceutical giants.
Why it matters: Anthropic hadn’t typically been as science-focused as Google and OAI, but that has changed in the past year — with a Life Sciences effort, this release, and a series of splashy hires (including Nobel winner John Jumper from DeepMind) pushing the lab deeply into the scientific arena for AI-enabled research and discovery.
QUICK HITS
🚀 Claude Sonnet 5 - Anthropic's newly released cheaper agentic model
🎥 Gemini Omni Flash - Google's video model for video generation and editing
🧪 Claude Science - Anthropic’s app for AI-powered scientific research
🐈 Longcat 2.0 - Meituan's open coding model trained on Chinese chips
OpenAI found a ‘compute multiplier’ that more than halves its inference costs, according to The Information, coming alongside the recent debut of its Jalapeño chip.
U.S. chipmaker Etched announced $800M in funding, while revealing a working inference chip with the full server rack as well as $1B in customer contracts.
AWS is committing $1B to a new Forward Deployed Engineering org, embedding thousands of engineers inside customer teams to accelerate agentic AI builds.
X rolled out a hosted MCP server, letting AI tools like Grok, Claude, and Cursor easily connect to the X API and read or take actions on the social media platform.
OpenAI teased a new hardware collaboration with Work Louder, believed to be a keyboard device for its Codex platform launching on July 15.
COMMUNITY
Every newsletter, we showcase how a reader is using AI to work smarter, save time, or make life easier.
Today’s workflow comes from reader Matt K. in Austin, TX:
“I'm preparing for my 2nd marathon, and ahead of race prep I wanted to lose a couple of pounds to ease the pressure on my joints. I started with MyFitnessPal, but they gated a lot of the features — most importantly macro tracking — and it didn't let me interact with the app in a natural way.
I realized I could use ChatGPT as my nutrition and performance coach. I uploaded my DEXA scan results, input my goals, timeline, and activity levels, and connected it directly to a Google spreadsheet to act as my backend source of truth.
2.5 months later, I've lost approx 4 pounds of pure fat and maintained nearly all my muscle (confirmed via a follow-up DEXA scan) while keeping my performance levels high. Having a coach to tell me what to eat at my next meals and how much has been a game-changer. What might cost someone hundreds a month only costs me $20, and I feel like I'm building habits and routines I'll continue for the foreseeable future.”
How do you use AI? Tell us here.
Read our last AI newsletter: Meta’s brain-reading AI leaves letters behind
Read our last Tech newsletter: WhatsApp’s ‘usernames’ reservation race
Read our last Robotics newsletter: Agility Robotics going public at $2.5B
Today’s AI tool guide: Create winning ads with Claude in one command
RSVP to next workshop on June 30: Master AI video editing
That's it for today!
See you soon,
Rowan, Zach, Shubham, and Jennifer — the humans behind The Rundown
