In partnership with
Good morning. It’s Friday, April 3rd.
Here’s a cool story: A guy from Los Angeles spent about $20,000 vibecoding a startup in his house. He used AI to write the code, make his website, handle customer service, and create ads. His company Medvi was a telehealth provider for GLP-1 weight loss drugs.
He ended up doing $400 million in sales last year by himself (although he hired his younger brother), and is on track to do $1.8 billion this year.
AI runs the entire company.
-Jeff
AI Breakfast
You read. We listen. Let us know what you think by replying to this email.
Turn AI into Your Income Engine
Ready to transform artificial intelligence from a buzzword into your personal revenue generator?
HubSpot’s groundbreaking guide "200+ AI-Powered Income Ideas" is your gateway to financial innovation in the digital age.
Inside you'll discover:
A curated collection of 200+ profitable opportunities spanning content creation, e-commerce, gaming, and emerging digital markets—each vetted for real-world potential
Step-by-step implementation guides designed for beginners, making AI accessible regardless of your technical background
Cutting-edge strategies aligned with current market trends, ensuring your ventures stay ahead of the curve
Download your guide today and unlock a future where artificial intelligence powers your success. Your next income stream is waiting.
Anthropic finds emotion circuitry in Claude, and it's now central to alignment
Anthropic is peeling back the layers of AI behavior, identifying functional emotion vectors in Claude Sonnet 4.5. These internal patterns, like urgency or composure, aren't sentient but do causally steer model outputs. This discovery is now central to how they monitor alignment and structure pretraining. Operationally, the focus is on adaptive agent harnesses. By using context compaction and subagents, developers can let Claude orchestrate complex tasks while keeping latency and costs low.
Anthropic is expanding into life sciences with a $400M Coefficient Bio acquisition, while its leaked Claude Code has been cloned 8,000+ times on GitHub, exposing model-control methods and threatening its $380B IPO. The company is developing Epitaxy, a Claude Code desktop UI, and Conway, a persistent agent platform integrating Chrome, webhooks, and extensions. Claude agents have also influenced markets, correctly predicting Eli Lilly’s FDA approval for Foundayo, sparking early $LLY trades in a $50B market. Read more.
Google debuts Gemma 4 open-weight models
Google DeepMind expanded its ecosystem with the Gemma 4 open-weight model family, featuring 2B and 4B edge models alongside 26B MoE and 31B Dense versions. Built on Gemini 3 technology and released under an Apache 2.0 license for the first time, these models support 256K context windows and multimodal processing.
Simultaneously, Google Vids integrated Veo 3.1 and Lyria 3 for video and music generation, introducing text-prompted AI avatars and direct YouTube publishing. Google AI Studio added Focus Mode for UI edits, while AI Pro storage increased to 5TB. To address security, DeepMind identified six exploit paths for autonomous agents, including content injection and poisoned memory. Google Stitch also launched a prompt enhancer. Read more.
OpenAI red-teamer releases Excalibur, a no-code LLM agent framework
OpenAI President Greg Brockman recently stated that the company’s reasoning models provide a direct ‘line of sight’ to AGI, even as internal models solved three Erdős mathematical problems with elegant proofs. To support developer security, an OpenAI red-teamer independently released Excalibur, a no-code framework using markdown files for agent orchestration and structured memory.
The company is also shifting its commercial strategy, introducing usage-based pricing for Codex seats and reducing ChatGPT Business costs to $20 annually. This follows a reset of Codex limits to ban abusive accounts and recover compute. In the public sphere, OpenAI acquired the TBPN media network to improve AI communication. Addressing broader oversight, CEO Sam Altman admitted that he ‘miscalibrated’ the mood of distrust toward AI and the government in the recent Pentagon deal. Read more.
Atlas 1 by Willow delivers real-time speech-to-text with human-assisted accuracy, outperforming OpenAI and Deepgram.
Tripo Studio is a text-to-3D model generator for artists, game developers, AR/VR modelers.
Cosyra lets you run Claude Code, Codex, OpenCode, and Gemini CLIs from a persistent cloud terminal on your phone.
Lightning V3 delivers real-time, multilingual, human-like speech with ultra-low latency and instant voice cloning for enterprise voice agents.
Mngr runs hundreds of Claude agents in parallel, orchestrating coding workflows with fast startup, real-time visibility, and open-source CLI control.
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.
