Hey hustlers,
Google is 9 days from its biggest event in years. Anthropic is chasing a trillion-dollar valuation. And OpenAI quietly dropped three voice models in one day that nobody is talking about enough.
Different companies, different bets. Same direction: AI is moving into every surface you interact with. Your screen, your wallet, your voice.
Let's take a look.
👾 WHAT'S NEW IN AI
1. Google I/O Is 9 Days Away and the Confirmed Lineup Is Genuinely Insane
Google I/O 2026 opens on May 19 and confirmed sessions include Gemini 4 scoring 84.6% on ARC-AGI2, a new Ironwood TPU pushing 42.5 exaflops, AI glasses built with Warby Parker, a brand new desktop operating system, and Gemini running inside Boston Dynamics' Atlas robot.
These aren't leaks. They're confirmed sessions. One writer called it "a company trying to rewrite every technology category simultaneously."
Why you should care: If even half of this ships as announced, Google just became the most aggressive AI company in the room. The glasses, the robot, the OS: every computing surface humans interact with is on the table at once.
May 19 is worth blocking off your calendar
2. Anthropic Is Chasing a $900 Billion Valuation. Nearly a Trillion Dollars.
Anthropic is asking investors to submit allocations within 48 hours for a roughly $50 billion funding round estimated to close within two weeks, targeting a valuation of about $900 billion. A board decision is expected in May, with a potential IPO as early as October 2026. The company that didn't exist four years ago would surpass OpenAI as the most valuable private AI company in the world.
Why you should care: Anthropic's annualised revenue went from $9 billion at end of 2025 to $30 billion by March 2026. No company in American history has scaled revenue that fast. This isn't hype money. This is investors betting on a company with real, accelerating revenue that could IPO before the year ends.
3. OpenAI Launched Three New Realtime Voice Models
GPT-Realtime-2 brings GPT-5-class reasoning to live voice conversations. GPT-Realtime-Translate handles real-time speech translation across 70+ input languages into 13 output languages. GPT-Realtime-Whisper does streaming transcription. All available via API now. Voice AI just became a serious developer platform overnight.
Why you should care: Last week xAI launched a voice model closing real sales calls. This week OpenAI dropped three voice models in a single day. The pattern is clear: every major AI lab is racing to own the voice layer.
If you create content, run a business, or build anything, voice AI just became something you can actually integrate without duct-taping five tools together. Translation across 70+ languages alone opens markets that were completely out of reach six months ago.
👾 THE GOOD STUFF
🔧 AI Tool: Cotypist
An AI autocomplete that learns your writing style and
finishes your sentences across every app. Not just Gmail or Notion.
Everywhere you type. Currently free in early access.
If you write a lot, this is worth grabbing before the paid plans launch.
🐙 GitHub: mintplex-labs/anything-llm (59.8k stars)
With OpenAI now available on AWS Bedrock and multiple clouds, this official Python SDK just became even more relevant.
It abstracts away the cloud routing complexity and gives you clean API access regardless of where OpenAI's models are actually running.
🎬 YouTube: Claude Code: Build Your First AI Agent
With Anthropic now doubling Claude Code rate limits and chasing a trillion-dollar valuation, this is the right week to actually understand what Claude Code does.
This tutorial walks you through building your first AI agent from scratch.
👾 TO READ
SkillOS: The AI Agent That Actually Gets Smarter
the More You Use It
📎 https://arxiv.org/abs/2604.24618
Most AI agents solve problems and move on. Ask the same thing tomorrow and they start from scratch. SkillOS tackles this by letting agents distill reusable skills from past interactions, using high-quality skill curation as the key lever for self-evolution
What they built: A system where agents don't just complete tasks but extract what they learned as reusable skills, rank those skills by quality, and apply them automatically to future tasks without being told to.
What they found: Agents using SkillOS consistently outperformed standard agents on long, complex task sequences because they stopped re-solving problems they'd already solved.
Why it's interesting: This week Anthropic raised billions to scale Claude agents and Google is about to ship Gemini inside robots. Neither is useful long-term if the agent forgets everything after each session. SkillOS is the research pointing toward agents that actually compound in value over time, the way a good employee does, not just a tool you have to re-brief every morning.
🧵 Thread Drop
Google wants to put AI in your glasses, your robot, and your OS. Anthropic wants a trillion dollars to make it happen faster. And OpenAI just made voice AI something developers can actually build real products with.
Three companies. Three very different moves. One very obvious destination.
👾 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.
