In partnership with
Hey {{first_name | there}},
So recently, Claude Code’s source code got accidentally leaked by Anthropic in a release.
By Wednesday morning, the code was mirrored across 8,100+ GitHub repositories. Anthropic issued a DMCA takedown. Then things got worse.
The takedown accidentally nuked their own public repositories and thousands of legitimate projects. Developers went ballistic on social media.
Anthropic retracted most of the notices within hours, but the damage was done.
Chinese mirrors on Gitee were already live, and the code is now permanently out there.
What the leak revealed:
KAIROS Dream Mode: an unreleased autonomous agent that reviews its own memories overnight, consolidates learnings, and responds to GitHub webhooks while you're asleep.
Your structured data is invisible: Claude code only processes the <body> tag. All your JSON-LD, Schema.org markup, and FAQ schemas in the <head> section? Ignored completely.
Token burn is real: Claude Code spends 30-50K tokens at the start of every conversation just exploring your codebase before doing any real work.
There's a hidden Tamagotchi: Developers found an unreleased AI terminal pet called “BUDDY” with 18 species, rarity tiers, and stats like CHAOS and SNARK.
Why it matters:
Coming off another leak just days earlier (the "Mythos" model), this makes Anthropic look careless at exactly the wrong time.
But what matters more is what got exposed: your AI coding assistant is burning 30-50K tokens per session on background exploration, your SEO schema markup is completely ignored, and there's a whole autonomous agent system running that nobody knew about. These tools cost more and do more than the marketing suggests.
Quick question:
What interests you the most?
🛠️ AI Tools Worth Checking Out
SeaOcean — Run a complete SEO audit with 75+ checks in seconds to improve your site performance.
Contextli — Convert speech into context-aware text for emails, Slack, and LinkedIn effortlessly.
PromptFix — Analyze and improve your AI prompts with scoring, suggestions, and A/B testing.
Blink.new — Build production-ready apps just by chatting with AI, no coding required.
OpenAI just closed the biggest fundraise in venture capital history. $122 billion at an $852 billion valuation. Amazon, Nvidia, and SoftBank led the round, and the company says it's merging everything into one unified superapp.
What you need to know:
OpenAI is now doing $2B in revenue per month. For context, they hit $1B total revenue within a year of launching ChatGPT. By end of 2024, they were at $1B per quarter. Now it's $2B monthly.
The superapp consolidates ChatGPT, Codex, browsing, and agent tools into one experience. Think less "collection of products" and more "single system that does everything."
Over $3 billion came from individual retail investors through bank channels. OpenAI also got added to ARK Invest ETFs so regular people can buy exposure.
Why it matters:
This isn't just about the dollar amount. OpenAI is betting that consumer familiarity becomes the trojan horse for enterprise adoption. You use ChatGPT at home, you push for it at work. The superapp strategy makes that easier by giving you one interface instead of juggling separate tools.
Oracle is laying off thousands of employees as it continues ramping up spending on AI data centers. The cuts are the largest in company history, coming as Oracle's stock is down 25% this year on concerns about debt and cash flow.
What you need to know:
Oracle has 162,000 employees. The layoffs are in the thousands but the exact number hasn't been disclosed
The company raised $50B in debt and equity in January to fund AI infrastructure buildout
Cutting 20K-30K employees could free up $8B-$10B in incremental cash flow, per TD Cowen analysts
Why it matters:
Oracle is smaller than AWS, Google Cloud, and Azure, but it's betting hard on AI infrastructure. The problem? Building data centers burns cash faster than the revenue comes in. These layoffs are Oracle trying to stay solvent while it waits for the AI boom to actually pay off. If demand holds, this works. If it doesn't, Oracle just torched its workforce for nothing.
OpenAI signed Smartly as its first adtech partner to help brands build interactive, conversational ads in ChatGPT. The ads will use real-time optimization and eventually let users chat with ads directly.
What you need to know:
OpenAI launched basic contextual ads in February for free and $8/month Go tier users
Smartly will help brands tweak ads in real time based on performance
The goal is conversational ad formats, like a chatbot that helps you pick a gift or compare products
OpenAI hit $100M in annualized ad revenue and is working with 600+ advertisers
Why it matters:
OpenAI could hit $25B in ad revenue by 2030 if this works, but the challenge is trust. Ads need to stay separate from ChatGPT's organic answers or users will bail. Google shows ads in AI Overviews but not in Gemini. Anthropic rejected ads entirely in Claude. OpenAI is betting it can thread the needle, but one misstep and the backlash will be brutal.
A Tool That Saves 50K Tokens Per Claude Code Conversation
A developer built ai-codex, a script that pre-indexes your codebase and generates 5 compact markdown files. It eliminates the 30-50K token exploration phase Claude Code does at the start of every conversation.
What it does:
Scans your project once and outputs routes.md, pages.md, lib.md, schema.md, and components.md
You add one line to CLAUDE.md telling Claude to read these files first
Future conversations skip the exploration phase entirely
Why it matters:
Claude Code burns tokens exploring your codebase at the start of every session. This is a hidden cost nobody talks about. The tool is open source, works with Next.js and TypeScript projects, and takes 5 seconds to run. If you're working on a large codebase, this is a no-brainer.
That's it for today.
If Anthropic's leak taught us anything, it's that these AI tools are way messier under the hood than the polished demos suggest.
What are your thoughts on all of this? Reply and let me know I read everything!
See you next time.
- Aashish
Become the go-to AI expert in 30 days
AI keeps coming up at work, but you still don't get it?
That's exactly why 1M+ professionals working at Google, Meta, and OpenAI read Superhuman AI daily.
Here's what you get:
Daily AI news that matters for your career - Filtered from 1000s of sources so you know what affects your industry.
Step-by-step tutorials you can use immediately - Real prompts and workflows that solve actual business problems.
New AI tools tested and reviewed - We try everything to deliver tools that drive real results.
All in just 3 minutes a day