Hey hustlers,
Google just got caught preparing something big before its biggest event of the year. Elon's AI quietly became one of the cheapest frontier models on the planet. And Anthropic just partnered with Wall Street to go after the consulting industry.
Three very different moves. All pointing at the same thing: the AI wars are moving from research labs into boardrooms, bank accounts, and your video feed.
Let's take a look.
👾 WHAT'S NEW IN AI
1. Google Accidentally Leaked Its Next Big AI Tool Before I/O 2026
A screenshot from inside Gemini's video generation tab surfaced this week with a previously unseen line: "Start with an idea or try a template. Powered by Omni."
Nobody at Google announced anything. The string just appeared. The leak suggests Google is bundling video and image generation into one consumer-facing model called Omni, a structurally different bet from OpenAI's standalone Sora. Right now Google runs separate models for everything: Veo for video, Nano Banana for images.
Omni would collapse all of that into a single product. Google I/O 2026 opens on May 19, which is two weeks away. The timing is not subtle.
Why you should care: If the leak holds up at I/O, Google is about to ship the first top-tier omni-model that handles video and images inside one system. For content creators, that means one tool to generate your thumbnails, your short-form clips, your B-roll, and your social posts. The current juggling act between apps is about to get a lot simpler, at least on Google's side.
2. xAI Completed the Full Rollout of Grok 4.3 and It's Aggressively Cheap
xAI completed the full rollout of the Grok 4.3 API, slashing input prices by approximately 40%, expanding the context window to 1 million tokens, and introducing native video input support for the first time.
The largest single benchmark improvement is on real-world agentic tasks, where Grok 4.3 surpassed Gemini 3.1 Pro Preview, Muse Spark, GPT-5.4 mini, and Kimi K2.5.
It also now generates formatted PDFs, spreadsheets, and PowerPoint decks directly from conversation. The new voice cloning suite prices voice AI at 1/10th of the industry standard, completely rewriting the cost curve for high-volume audio scenarios like customer service bots and in-car voice assistants.
Why you should care: This pricing strategy means xAI has effectively rewritten the cost model for agentic applications. If you've been using Claude or ChatGPT for heavy document work or long agentic sessions, Grok 4.3 at these prices is worth a serious look. The catch: still no persistent memory between sessions at $300/month. That's a real gap compared to ChatGPT and Claude.
3. Anthropic Just Declared War on the Consulting Industry With $1.5 Billion Behind It
Anthropic is closing a joint venture with Blackstone, Goldman Sachs, and Hellman & Friedman to sell Claude directly to private equity-backed companies. Wall Street is officially buying into AI at a corporate infrastructure level.
Hours before the announcement, Bloomberg reported that OpenAI was launching a near-identical venture called The Development Company. Both companies announced the same play on the same day.
Why you should care: For every dollar companies spend on software, they spend six on services. That ratio has made consulting a multitrillion-dollar industry and AI-native firms are now positioning to disrupt it. Anthropic and OpenAI are not just selling subscriptions anymore. They're coming for McKinsey's clients. The AI economy just got a lot more interesting.
👾 THE GOOD STUFF
🔧 AI Tool: Gumloop
An AI automation platform built for content teams and agencies.
You describe what you want in plain English and it builds the workflow,
from drafting newsletters to generating reports to pulling data across tools.
No coding, no Zapier-style box dragging. Free tier available.
Getting a lot of traction in creator and media workflows right now.
🐙 GitHub: anthropics/claude-cookbooks (42.2k stars)
Anthropic's official collection of code recipes showing exactly
how to build real things with Claude - agents, document analysis, tool use, and more.
Just updated with new examples this week.
With Anthropic now deploying Claude inside Goldman Sachs and Blackstone
portfolio companies, this is the playbook they're using.
🎬 YouTube: Claude Code: Build Your First AI Agent
With Anthropic now partnering with Wall Street to deploy Claude
inside real companies, understanding how Claude Code actually works
is worth your time.
This tutorial walks you through building your first AI agent from scratch.
👾 TO READ
Thinking with Visual Primitives — DeepSeek
📎 https://arxiv.org/abs/2604.24618
Most AI models can look at an image. But they still mostly reason in words.
Ask GPT-5.4 to count people in a dense crowd photo and it will likely get it wrong, because language is a clumsy tool for tracking spatial things.
What they built: A framework where AI models don't just describe what they see, they literally point at it. Instead of saying "the person on the left," the model drops a coordinate marker on the actual pixel. Points and bounding boxes become part of the thinking process, not just the output.
What they found: Despite using smaller models and fewer image tokens, this approach matches or surpasses GPT-5.4, Claude Sonnet 4.6, and Gemini 3 Flash on challenging visual reasoning benchmarks including counting, spatial deduction, and maze navigation.
Why it's interesting: Google just leaked a video model called Omni that handles images and video in one system. This paper shows the other side of that story: even with better video models, AI still struggles to reason spatially once it sees something. Teaching AI to point, not just look, might be the fix. DeepSeek did it with a smaller model. That's the quietly important part.
🧵 Thread Drop
The week wasn't about models getting smarter.
It was about AI getting deployed into money, media, and markets in ways that are hard to reverse.
That's the shift worth paying attention to.
👾 See you soon 👾
Want to work together?
If you're building something in AI and
think our audiences would vibe with it, let's talk.
