Grow your business with AI tools and hacks?
Earn more, work less with AI.
Stay up-to-date with the latest AI tools, hacks, and prompts delivered straight to your inbox.
Want to sponsor a newsletter? Find out more.Plus: Gemini 4 Carbon goes public, ChatGPT starts shopping for you, and Claude shows up to your meetings
Sponsored by
Generate high-quality audio with our AI voice generator for audiobooks, videos, and podcasts
Try it nowThis was the week the AI industry learned that its single biggest risk isn't a benchmark or a bubble — it's one guy's calendar. Jensen Huang is out at Nvidia, Congress is suddenly very interested in where AI chips end up, and every hyperscaler CFO spent the week quietly rewriting a slide deck.
Meanwhile, the models didn't wait around for the drama to settle: Gemini 4 Carbon went fully public, Anthropic shipped Claude into your video calls, OpenAI gave its agents a wallet, and DeepMind used AI agents to formally verify the Signal protocol — a problem cryptographers have been circling for a decade.

On this day in 2025, the AI world was in the eye of the hurricane: OpenAI had just shipped GPT-5.1 with its "Instant" and "Thinking" modes days earlier, and Google was about 72 hours away from dropping Gemini 3 on everyone's heads. Week 47 has a history of being spicy, apparently.
🛡️ Congress wants to tighten export rules on AI chips after Nvidia spy scandal
The fallout from Jensen Huang's exit is already turning into policy. Lawmakers on both sides of the aisle are lining up behind bills that would clamp down on advanced chip exports, and several have demanded that Huang and other Nvidia leaders testify. Three export-control proposals are currently riding along with the defense authorization bill. Translation: the era of "light-touch regulation because Jensen said so" is over, and every AI company that budgeted on cheap, abundant GPUs now has a policy risk line item.
🪨 Gemini 4 Carbon goes public — and undercuts Claude on API price
Google's newest frontier model is now broadly available on the paid Gemini API and Gemini for business, debuting at the top of Artificial Analysis's general intelligence index and beating Fable 5.2's best listed run on Terminal-Bench 4.0 (65.2% vs 61.6%). The catch: several of Google's coding claims can't be verified yet, since Carbon is simply missing from the public CursorBench and WANDR tables. Still, a frontier model that's both top-ranked and cheaper per token is going to move a lot of production traffic.
🛒 ChatGPT's new Wallet lets AI agents shop and check out for you
OpenAI launched a wallet feature that lets its "dots" agents pick products and complete the purchase on your behalf — initially limited to some U.S. ChatGPT Pro subscribers with a Visa, Mastercard, or Amex. Every checkout needs a Face ID or fingerprint approval, with spending limits and category pre-approvals coming later. The interesting subplot: Amazon is still blocking AI agents outright, so agentic commerce is becoming a retailer-by-retailer land grab.
🧑💻 Anthropic launches Claude Meetings, Whiteboards, and Sheets
Claude can now join a video call as a visible participant, follow the conversation, draft the plan being discussed, sketch a timeline on a whiteboard, answer questions in chat, and pull last year's numbers out of your connected apps — all while people are still talking. Add Whiteboards (beta on paid plans) and Sheets, and Anthropic has quietly assembled an office suite. OpenAI shipped something similar weeks ago. Microsoft and Google are the target.
🔐 DeepMind's AlphaProof Code formally verified the Signal protocol
Google DeepMind guided AI agents to mathematically prove key security properties of the actual code behind the Signal protocol — the thing protecting billions of conversations in Signal and Google Messages. It's the first time a widely deployed cryptographic protocol has been formally verified at this depth by any method, AI-assisted or not. Formal verification has always been the gold standard nobody could afford; agents just made it cheaper.
⚡ addyosmani/agent-skills: Addy Osmani's repo of senior-engineer workflows for AI coding agents is back on GitHub Trending, adding 788 stars in a single 24-hour window on Nov 11 and pushing past 115,700 stars with 12,100+ forks. It packages real process into six slash commands — /spec, /plan, /build, /test, /review, /ship — plus /webperf and /code-simplify, and works across 70+ agents including Claude Code, Cursor, Codex, and Copilot. The signal here is bigger than the stars: teams have stopped chasing raw model capability and started buying guardrails.
🔌 simonw/llm: Simon Willison pushed datasette-llm 0.1a11 this week, continuing his quiet campaign to make LLMs a boring, composable layer in the data stack rather than a product. The underlying llm CLI and library remain the best way to script dozens of models — local and hosted — behind one interface, with plugins for embeddings, tools, and templates. If you're building internal eval or data-labeling pipelines, start here before you reach for a framework.
Coming in Hot
Open-source email platform for SaaS teams
Launch platform for makers to showcase products
AI design agent for e-commerce product listings
AI visibility tracking with white-label marketing reports
Unified social media data API for developers
Humanize AI text to sound natural and undetectable
AI website builder for busy startup founders
AI email marketing built for SaaS founders
AI agents that grow your pipeline across every location
AI image generator that thinks before creating
Submit your AI tool to the #1 AI tools marketplace.

What it does: Inbylo combines AI-powered customer messaging, social media scheduling, graphic design, and team collaboration in one workspace. It helps small businesses and agencies manage Facebook, Instagram, and WhatsApp conversations while automating common customer replies.
Why It's Worth Your Time
Pricing
Why it matters: Most small businesses use separate tools for social media messages, chatbot automation, post scheduling, and graphic design. Inbylo brings these workflows together in one affordable platform. Its focus on AI customer replies and agency-friendly workspaces makes it particularly useful for Facebook and Instagram-based businesses, although its limited social network support may not suit larger teams managing multiple channels.

What it does: Recto gives AI coding agents something they've always lacked — eyes and a canvas. Instead of describing a layout in a prompt and hoping, you (and your agent) work on a visual surface to create and edit websites, UI, animations, and 3D designs, then push the result into real code.
Why It's Worth Your Time
Pricing
Why it matters: Coding agents got genuinely good at logic and stayed genuinely bad at taste. Tools like Recto are the correction — a shared visual layer where humans judge and agents execute. If 2026 was the year agents wrote the code, 2027 is the year someone finally makes them care how it looks.
🔥 Meta shipped Muse Ember, its cheapest and lightest model yet, built for the grunt work agents repeat all day — tool calls, routing, extraction, long-document parsing. At 10 cents per million input tokens it undercuts Haiku and Luna and is cost-competitive with open-source flashes. The pitch: "Muse Wonder plans, Muse Ember handles the volume."
🤖 Gemini Robotics 2.5 hit a 96% success rate packing automotive parts in real-world tests at Mercedes-Benz, completes tasks 20–30% faster than its predecessor, and ships with a new jailbreak benchmark for physical steerability. Robot foundation models are quietly getting boring-reliable.
🐱 Meituan open-sourced LongCat-2.5 under MIT — a 1.6-trillion-parameter multimodal agent model with native vision, trained entirely on Chinese hardware and deployed on Huawei Ascend chips. It scores 88.4 on Terminal Bench, just behind Claude Opus 5. Also 3.56 TB in BF16, so maybe don't try this at home.
💰 LangChain raised $300M at a $3B valuation, more than doubling in a year. Its pitch has shifted entirely from "framework" to "let businesses own their own intelligence" — 7,000 LangSmith customers now including Nvidia, Toyota, and Salesforce.