Sponsored by

Good morning. It’s Wednesday, March 27th.

Now that Elon’s lawsuit against OpenAI was thrown out by the courts, OpenAI is preparing to go public (possibly filing as early as Friday) for around $730B.

However Polymarket is showing only 70% odds that the stock will be on sale to the public by December 31st… interesting.

-Jeff
AI Breakfast

You read. We listen. Let us know what you think by replying to this email.

Stop Losing Your Money. It's time to upgrade your trading platform.

Your current trading platform is probably letting you down

  • Limited assets (no international stocks, no commodities, no pre-IPO companies)

  • Limited ability to short

  • Limited access to leverage

  • Limited trading hours

Liquid is one of the fastest growing trading platforms, allowing users to trade stocks, commodities, FX, and more 24/7/365 from their phone and computer.

Trading on Liquid is as simple as:

  1. Pick an asset

  2. Pick long or short

  3. Pick your position size and leverage

  4. Place your trade

The best part is that Liquid markets never close. So no matter what is going on in the world, you are able to keep your portfolio positioned properly.

Anthropic now grades your prompting

Anthropic's experimental Claude Mythos system recently attempted to solve the Erdős unit-distance conjecture, a combinatorial geometry problem first proposed in 1946.

Rather than relying on a single reasoning path, the system utilizes a multi-agent framework where isolated Claude Code instances explore multiple hypotheses in parallel. An aggregator instance then compiles these separate threads into a candidate proof using a geometry-of-numbers approach. While some mathematicians noted the output was less rigorous than a recent milestone from OpenAI, the Mythos system achieved a similar solution through a more direct structure.

In tandem with this technical development, Anthropic is introducing an AI Fluency scorecard within the Claude interface to evaluate user interaction patterns. The tool, developed from an analysis of approximately 10,000 interactions, monitors 11 behavioral signals across chat, coding, and collaboration workflows. By assessing metrics such as prompt design, iteration quality, and verification habits, the system generates a composite score designed to measure and track user AI literacy.

Co-founder Chris Olah recently addressed the Vatican during the release of the papal encyclical Magnifica humanitas. Olah noted that modern neural networks are grown rather than engineered, resulting in complex internal representations that resemble certain aspects of human cognition. However, the encyclical adopted a stricter position and explicitly ruled that autonomous systems must not be permitted to dictate moral decisions. Read more.

A developer used Codex to rewrite firmware and upgrade a budget MP3 player

An independent developer, known online as bunkaich, successfully upgraded a budget MP3 player from AliExpress by using Codex to rewrite its firmware. Frustrated by persistent Bluetooth audio stuttering and clunky menu navigation, the developer disassembled the device to expose its internal microchip.

By feeding images and specifications of the chip into Codex, the AI independently investigated the hardware architecture. Under the guidance of the AI, the developer connected the device to a Mac, extracted the original operating system, and allowed Codex to analyze the binary files. Codex identified the root causes of the system flaws and generated a custom OS.

The new firmware completely resolved the Bluetooth audio stuttering. It also restructured the user interface, switching the menu navigation from counterintuitive left-right button inputs to a logical up-down selection system. The developer noted that the success relied partly on luck due to the original OS structure. Read more.

xAI triples parameter count: Grok V9-Medium expected to launch within weeks

xAI has rolled out Grok Build in Beta for SuperGrok and X Premium+ users, pairing image and video generation with a CLI designed for building automations and orchestrators. The platform introduces a specialized Plan Mode, expanding Grok from a basic chatbot into a more functional development assistant.

Alongside the software rollout, Musk announced that Grok V9-Medium has finished training. The new foundation model scales up to 1.5 trillion parameters, a massive jump from the 0.5-trillion parameter V8-Small model currently handling production traffic. To sharpen its performance on complex programming tasks, xAI fed significant amounts of Cursor data into the model during supplementary training.

The model is now entering fine-tuning and reinforcement learning, with a public release expected in two to three weeks. Read more.

Frontier models and product moves

Agents and the agentic stack

Business, labor, and institutions

Hardware and infrastructure

Research and benchmarks

Developer tools and open source

Media, audio, and creative AI

Robotics

Watch

Bond prompts an AI GTM engineer to turn raw buyer signals into fully orchestrated outbound campaigns.

Rezonant turns messy product ideas into codebase-grounded technical specs and structural tasks optimized for human engineers and coding agents.

Crade reads your screen to answer questions, solve errors, and complete everyday tasks right where you work.

Brew uses natural language to spin up on-brand email campaigns, code logic, and cross-client responsive designs instantly.

Dobs AI automates financial due diligence, parsing messy data rooms into fully traceable workbooks with live formulas.

Thank you for reading today’s edition.

Your feedback is valuable. Respond to this email and tell us how you think we could add more value to this newsletter.

Interested in reaching smart readers like you? To become an AI Breakfast sponsor, reply to this email or DM us on X!

Thinking of starting your own newsletter? AI Breakfast readers who sign up with Beehiiv receive a 14-day free trial and 20% off for 3 months.

Keep Reading