Claude Max Limits, DeployCo, Codex on IOS | Weekly Digest
PLUS HOT AI Tools & Tutorials
Hey! Welcome to the latest Creators’ AI Edition.
Google dropped its biggest hardware announcement of 2026 the week before I/O — a new laptop category called Googlebook, built from the ground up for Gemini. Anthropic went the other direction and embedded Claude directly into QuickBooks, PayPal, and HubSpot, targeting the 44% of the US economy that enterprise AI has ignored. Meanwhile the Musk v. OpenAI trial wrapped closing arguments May 14 — the jury deliberates Monday, with Altman's future hanging in the balance. Today we have:
Featured Materials 🎟️
News of the week 🌍
Useful tools ⚒️
Weekly Guides 📕
AI Meme of the Week 🤡
AI Tweet of the Week 🐦
(Bonus) Materials 🎁
Keep your mailbox updated with practical knowledge & key news from the AI industry!
Featured Materials 🎟️
Anthropic Puts Claude Inside QuickBooks, PayPal, and HubSpot — for Free 🏪
Small businesses make up 44% of US GDP and employ nearly half the private-sector workforce. Their AI adoption has consistently lagged enterprises. Anthropic launched Claude for Small Business on May 13 to close that gap — not with a new chat interface, but with prebuilt workflows inside the tools small business owners already pay for.
The product runs through Claude Cowork and connects to QuickBooks, PayPal, HubSpot, Canva, DocuSign, Google Workspace, and Microsoft 365 via a toggle install. No IT team, no custom integration. Fifteen workflows ship out of the box: payroll planning, invoice chasing, month-end close, cash-flow forecasting, sales campaigns, contract routing, customer service. Every action requires owner approval before Claude sends, pays, or posts anything.
The business case:
Anthropic president Daniela Amodei was direct: “Small businesses make up nearly half the American economy, but they’ve never had the resources of bigger companies. AI is the first technology that can finally close that gap.” There’s no extra charge beyond existing Claude licenses and whatever software the business already uses. Anthropic is also launching a free 9-lecture AI Fluency course co-developed with PayPal, plus a 10-city US roadshow starting May 14 — Chicago, Tulsa, Dallas, New Jersey, Baton Rouge, Birmingham, Salt Lake City, Baltimore, San Jose, Indianapolis. 100 small business owners per city, half-day live workshop.
Why this matters beyond SMB:
Dario Amodei warned at Anthropic’s Financial Services event last week that SaaS companies that don’t adapt will go bankrupt. He wasn’t talking about small businesses — he was talking about QuickBooks, HubSpot, DocuSign. Claude for Small Business is Anthropic embedding itself one layer below those products. If it works, the leverage shifts.
Enterprise AI was always going to reach small businesses eventually. Anthropic just decided not to wait. The HVAC company and the 30-person real estate brokerage are now the target — and the SaaS tools they already use are the distribution channel.
Source: Anthropic
Anthropic Changes the Rules for Claude Power Users 💳
Anthropic quietly restructured the economics of its paid plans this week — and the developer community noticed immediately.
Every paid Claude plan now includes a monthly API credit equal to the dollar amount of the subscription for programmatic usage: the Agent SDK, claude -p, GitHub Actions, and third-party harnesses like OpenClaw. Pay $200/month, get $200 in API credits on top of your interactive Claude.ai and Claude Code usage.
On paper, a good deal. In practice, power users who had been running third-party Claude harnesses at an estimated 70–90% discount from standard API pricing called it a rug pull. The backlash was immediate — Theo, Jeremy Howard, Matt Pocock, and Omar Sanseviero all weighed in publicly within hours of the announcement.
Anthropic partially offset the reaction with a separate move: +50% to Claude Code weekly limits through July 13, stacked on top of the previously announced 2× increase to the 5-hour limit. Claude Code users on interactive plans are getting more. Third-party harness users are getting metered.
What it signals:
Claude Code now has enough brand and market position that Anthropic is putting its most favorable pricing behind its own tools. The subsidy era for alternative harnesses is ending. Meanwhile OpenAI — the challenger in this specific battle — is going the other direction: two free months of Codex for enterprise customers who switch in the next 30 days.
Anthropic built its developer base partly on pricing generosity toward third-party tooling. This week it started converting that goodwill into margin. The question is how many power users follow Codex.
Source: X
Share this post with friends, especially those interested in AI!
News of the week 🌍
Musk v. OpenAI: Closing Arguments Done, Jury Deliberates Monday ⚖️ — The first phase of the trial wrapped May 14 after closing arguments from both sides. Musk’s team argued OpenAI “misused millions in donations” and betrayed its nonprofit mission. OpenAI countered that Musk’s $38M came without strings and that the mission continued through structural changes. The nine-person jury begins deliberating Monday, May 18 — but their verdict is advisory only. Judge Gonzalez Rogers makes the final call on liability, and a separate remedies phase (potential Altman removal, unwinding the 2025 recapitalization) runs concurrently before the judge alone.
Anthropic in Talks to Raise $30B at $900B+ Valuation 💰 — Bloomberg confirmed on May 12 that Anthropic is in early discussions to raise at least $30B, which would triple its February valuation of $380B and put it above OpenAI’s $852B post-money figure. The round is expected to close by end of May and may be Anthropic’s last private round before an IPO targeted for October 2026. ARR is reportedly at $14B, with Claude Code alone generating $2.5B annually.
Amazon Replaces Rufus With Alexa for Shopping 🛒 — Amazon launched Alexa for Shopping on May 13, retiring the Rufus brand and combining Rufus’s product expertise with Alexa+’s personalization layer. Available free to all US customers — no Prime, no Echo device required. It tracks price history, schedules recurring orders, and includes a “Buy for Me” feature that purchases on your behalf at other retailers. 300 million customers used Rufus in 2025.
OpenAI Brings Codex to Mobile — Manage Your Coding Agent From Your Phone 📱 — OpenAI pushed Codex into the ChatGPT mobile app on iOS and Android. Start a coding agent session on your machine, then manage it from your phone: review diffs, approve commands, inspect terminal output, look at screenshots, switch models. No desk required. The gap between "AI coding agent" and "always-on background worker" just got a lot smaller.
OpenAI Launches $4B Deployment Company — Anthropic Mirrors the Move 🏢 — OpenAI stood up a standalone enterprise consulting arm on May 11 with $4B from TPG, Goldman Sachs, McKinsey, Bain Capital, and 15 others — embedding Forward Deployed Engineers directly inside Fortune 500 clients, Palantir-style. Acquired London firm Tomoro (150 engineers, clients include Tesco and Virgin Atlantic) to launch with an existing book of business. Same day, Anthropic announced its own $1.5B deployment JV with Blackstone, Hellman & Friedman, and Goldman Sachs. The model race is over. The deployment race is on.
Google Announces Googlebook — A Laptop Built Around Gemini 💻 — Google unveiled a new laptop category on May 12 at the Android Show: I/O Edition. Googlebooks ship with Gemini Intelligence built into the OS — Magic Pointer cursor, Android app streaming from phone, vibe-coded widgets on demand, seamless file sync. Manufacturing partners: Acer, Asus, Dell, HP, Lenovo. First units land this fall. The kicker: a federal antitrust judge had just barred Google from paying for AI default placement on third-party devices. Google's answer was to build its own.
Share this post with friends, especially those interested in AI!
Useful tools ⚒️
⭐ Spellar 3.0 — Most meeting tools give you notes. Spellar 3.0 gives you memory. It joins your calls without a bot, captures every word, and builds persistent context across all your meetings — so you can ask what a client said three calls ago or surface decisions from last month. Runs on Claude, GPT, Gemini, or Perplexity. Exports to Notion, Jira, Linear, Google Docs. Mac, iOS, and web. Free tier available.
Higgsfield Supercomputer — A cloud-native AI agent that runs your entire content pipeline from one prompt. Tell it “build a week of Instagram ads plus competitor analysis” and it plans the work, picks the right models, generates the assets, and delivers them. Three layers of memory, 40+ built-in tools, connects to Slack, Notion, Figma and 30+ services. Learns your brand guidelines over time. Free to start.
Latitude for Claude Code — See exactly where Claude Code is burning tokens and why. Shows token usage by task, flags which prompts hit your limits most, and helps you optimize before you run out of budget mid-session. Built specifically for Claude Code users.
Blaze 2.0 — AI marketer for small and mid-size businesses: strategy, content, and ads in one tool. Describe your business, Blaze builds a content calendar, writes social posts, and creates ad copy — all in your brand voice. Built for founders who don’t have a marketing team.
Weavable — Context infrastructure for AI agents. Instead of flooding your agents with raw API data, Weavable maintains a continuous updating changelog across your work tools — HubSpot, Slack, Zendesk, Jira, GitHub, email — and serves agents the specific signals they need through a single MCP endpoint. One-tenth the tokens compared to direct app connections, according to their own evals.
Weekly Guides 📕
How AI Agents Are Built in May 2026 — Step-by-step guide to building your first real AI agent from scratch: how to structure AGENTS.md, TASK.md, and MEMORY.md files, how the reasoning loop works, when to use tools vs. subagents, and how to avoid the most common failure modes.
Generate AI Videos From Claude With Higgsfield MCP — Step-by-Step — Higgsfield's own setup guide for connecting their MCP server to Claude in under a minute: how to add the custom connector, set permissions, and start generating 4K video from a plain-language brief.
Claude Code Security: MCP Connections, Permission Architecture, and Best Practices — Official Anthropic documentation on securing Claude Code in production: permission-based architecture, MCP server allowlisting, prompt injection defenses, sandbox configuration, and team security settings. Essential for any org deploying Claude agents at scale.
Building for the Gemini Intelligence System on Android — Developer’s Cut — Google’s official developer breakdown from the Android Show on May 12: how to integrate AppFunctions API, how task automation works across form factors, what’s available now vs. what’s coming at I/O, and how Googlebook changes the target surface area for Android apps.
AI Meme of the Week 🤡
AI Tweet of the Week 🐦
Bonus Materials 🎁
Cerebras IPO Opens +89% Above Price, $106B Valuation on Nasdaq Debut — The AI chip company that makes wafer-scale processors debuted on Nasdaq on May 14, opening at $350 vs. its $185 IPO price and raising $5.55B. Amazon and OpenAI are customers. The largest US IPO of 2026 so far, and a signal that AI infrastructure investing hasn’t cooled.
TikTok Symphony Creative Studio Now Open to All Businesses — ByteDance’s AI ad creation tool went generally available this week for every business on TikTok. Type a brief, get a polished video: synced audio, avatar in one of 30 languages, product-consistent visuals powered by Seedance 2.0. Every clip ships with an AI-generated label. For any brand spending on TikTok, this changes the cost floor for video ads overnight.
OpenAI Explores Legal Options Against Apple — Reuters reported May 14 that OpenAI is looking at potential legal action against Apple, which has been building its own AI features into iOS 27 using Anthropic and Google — not OpenAI. The story is early-stage but reveals a new front in the AI platform wars opening up just as the Musk trial was wrapping.
US and China Are Discussing AI Guardrails for the Most Powerful Models — Treasury Secretary Bessent confirmed May 14 that the US and China are in active talks about guardrails specifically for frontier AI models. The first public confirmation of bilateral AI safety talks at this level. Context: China’s four open-weight coding models dropped in a 12-day window last month, closing the capability gap faster than expected.
If you missed our previous updates, don’t worry, here they are:
Your take: Anthropic is pulling back pricing subsidies for third-party Claude harnesses — are you staying, or is this the moment you move to Codex? Drop it in the comments 👇