In partnership with

Welcome back!

Claude Fable 5 has arrived as the public version of the highly restricted Mythos, and it’s claiming the top spot across major coding, reasoning, and agent benchmarks. Early users say it can tackle remarkably complex work. Let’s unpack…

Today’s Summary:

• 🚀 Claude Fable 5 tops benchmarks

• 🍎 Apple rebuilds Siri with Gemini

• 🗣️ Google enables live voice translation

• 💻 MiniMax M3 adds 1M token memory

• 🤖 Gemma runs multimodal AI locally

• 👤 ChatGPT replaces Memory with profiles

• 🛠️ 2 new tools

TOP STORY

Anthropic launches Claude Fable 5

The Summary: Anthropic launched Claude Fable 5, the public version of its restricted Mythos 5 system. It ranks #1 across every major intelligence, coding, and agent benchmark. Built-in guardrails redirect cybersecurity, biology, and chemistry queries to Claude Opus 4.8. Early testers report it can complete very complex tasks with minimal guidance.

Key details:

• In Agent Arena real-world agentic tasks, Fable posted the widest confirmed task success margin ever recorded

• Stripe used it to complete a 50M-line Ruby migration in one day

• Requires 30-day data retention for safety monitoring

• Access is included on Pro, Max, Team, and Enterprise plans through June 22; after that, usage credits will be required

Why it matters: Fable arrives at the top of nearly every major benchmark that matters for coding, reasoning, vision, or agentic work. More importantly, it extends Anthropic’s lead at a moment when many expected frontier progress to become harder to achieve. Early users are sharing remarkable results from Fable, alongside complaints about its cost and restrictions.

System Card

FROM OUR PARTNERS

ChatGPT Agents You Can Use Today

22 ChatGPT Agents Built for Every Marketing Job

Most marketers use ChatGPT to do general research and then call it an AI strategy. The ones outperforming them are deploying specialized agents built for specific jobs.

We put together 22 plug-and-play ChatGPT marketing agents that handle the work eating your week, each with built-in instructions and structured outputs ready to go in under 5 minutes.

Subscribe to Marketing Against the Grain and get all 22 free.

Inside you'll find:

• Competitive intelligence agent that visits competitor websites and builds detailed comparison matrices automatically

• Customer feedback analyzer that ranks improvement opportunities by business impact

• Social listening specialist that monitors brand mentions and flags reputation risks before they escalate

• Campaign optimization agents that handle attribution analysis and surface what is actually driving results

Your competitors are already running agents like these.

Get 22 ChatGPT Marketing Agents free when you subscribe to Marketing Against the Grain today.

Get The Guide

APPLE

Apple rebuilds Siri around Google’s AI stack

The Summary: Apple is relaunching Siri with a new AI architecture that blends Apple models, Google Gemini, and a privacy-focused orchestrator. The assistant can understand personal context from emails, messages, calendars, photos, and on-screen content, then act system-wide and across apps in response to user requests. The core message is simple: Siri now performs what Apple previously promised.

Key details:

• New models were trained for Apple Silicon using Gemini

• Siri can now search years of personal history. Ask for a vacation photo from years ago, a buried email, or details from an old note

• Safari gains AI-powered page monitoring that can track changes

• Requires 12GB RAM with devices such as iPhone 17 Pro and M3 Mac

• Begins rolling out this fall in English-speaking markets. iPhone and iPad users in the EU will not get Siri AI due to regulations, although macOS and visionOS users in the EU will get it.

Why it matters: By using Gemini, Apple admitted it may not need to build its own frontier model, since it already controls the device and the OS. The system-wide index of personal data with privacy protection sounds like a solid approach. Many power users came away with a modest reaction, since it is mostly that Siri finally appears able to perform tasks Apple already demonstrated.

Features and architecture

GOOGLE

Google brings live voice translation to 70+ languages

The Summary: Google has launched Gemini 3.5 Live Translate, a speech-to-speech model that translates conversations in near real time across more than 70 languages. The system preserves tone, pacing, and pitch while generating translated audio only a few seconds behind the speaker. It is rolling out to Google Translate, Google Meet, and the Gemini Live API.

Key details:

• Enables more than 2,000 language pair combinations in Google Meet

• Unlike turn-based systems, it translates continuously while a person is speaking, balancing speed and accuracy in real time

• Every translated audio stream carries a SynthID watermark to detect AI-generated speech

• Rolling out globally in the Google Translate app on Android and iOS with wider Google Meet rollout later

Why it matters: Most translation tools force conversations into a stop-and-wait rhythm. Google’s approach targets the timing problem in live interactions. Language translation is becoming fast enough to fit into normal conversations. That will allow broader use in meetings, travel, customer support, or live events.

QUICK NEWS

Quick news

• MiniMax M3 pairs frontier coding with million-token memory

• Gemma 4 12B runs audio and video AI locally

• ChatGPT replaces Memory lists with User Profiles

TOOLS

🥇 New tools

• Codex Plugins - Skills for analytics, sales, design, banking

• Ideogram 4.0 - Open-weight 2K image model built for design