AI Daily: OpenAI Secures Massive Funding, Claude Code Source Leak, and New AI Updates

2026 Tech Spotlight: OpenAI Secures Billions in Funding, Claude Code Accidentally Leaks Developer Secrets

The world of AI has been hit by another bombshell. OpenAI has redefined market expectations with a staggering funding round, while Google and Ollama have introduced cost-effective solutions for video generation and local computing performance, respectively. Furthermore, the accidental leak of Claude Code’s source code offers a rare glimpse into the authentic and humorous daily lives of a top-tier development team. This article provides a comprehensive analysis of these trending tech topics.

To be honest, the tech world has something new every day, but today’s news is particularly striking. While giant corporations are pouring billions into massive computing architectures, some top development teams are quietly raising digital pets in their terminals. This coexistence of extreme commercialization and profound humor is exactly what makes the tech industry so fascinating. Let’s break down today’s highlights one by one.

OpenAI’s Billion-Dollar Funding Secured, Ambitions for a Superapp

When it comes to infrastructure expansion, OpenAI has undoubtedly dropped another bombshell. The company just announced the completion of a $122 billion funding round, bringing its post-money valuation to an incredible $852 billion. This capital will directly push the limits of computing power. One might wonder where all this money is going. The answer is clear: to build a ubiquitous intelligent system.

Currently, ChatGPT’s weekly active users have surpassed the 900 million mark, with over 50 million subscribers. Did you know? Their monthly revenue is as high as $2 billion. This is truly an incredible milestone. With such massive capital support, OpenAI is actively working to integrate ChatGPT, Codex, and web browsing features into a single, unified “Superapp.”

This is definitely more than just a UI update. Powered by the GPT-5.4 model, future systems will be able to more accurately understand user intent and execute complex tasks across platforms. More computing resources lead to smarter models, which in turn attract more users. This simple yet powerful flywheel effect continues to reshape the working habits of businesses and the general public worldwide.

Google Veo 3.1 Lite Debuts: The King of Value in Video Generation

While the market’s attention is focused on massive funding, Google has chosen to focus on practicality and cost control. The cost of video generation has always been a major pain point, often deterring many small creators. Google’s latest release of Veo 3.1 Lite addresses this issue precisely.

What makes this new video generation model so special? It retains the exact same generation speed as Veo 3.1 Fast, but the operational cost has been slashed by over 50%. This means developers can build high-compute visual applications without the financial pressure. Whether you need a 16:9 or 9:16 aspect ratio, or even high-definition output at 720p and 1080p, Veo 3.1 Lite handles it with ease.

This model is already available to the public via the Gemini API and Google AI Studio. Notably, Google also teased a price reduction for Veo 3.1 Fast on April 7th. This two-pronged pricing strategy will undoubtedly attract more people to integrate visual generation technology into their daily workflows.

Claude Code Source Leak: Revealing Developer Humor and Struggles

The tech world is always full of surprises. Often, the code left behind by engineers is more entertaining than official press releases. A leaked source code of Claude Code recently sparked heated discussions on community forums.

Keen-eyed users discovered that the Anthropic team had actually built a “Tamagotchi” system called /buddy into the terminal. In fact, according to the leaked source code, the salt for this system’s parameters was set to “friend-2026-401,” which is an April Fool’s Day easter egg prepared by the development team for April 1st. By entering a command, users can hatch their own ASCII pets, including capybaras, dragons, ghosts, and even a mysterious creature called “chonk.” To avoid strict internal code scanners, they even converted the pet name “duck” into hex code; and next to the random number generator for the pet system, an engineer left a humorous note: “This algorithm is enough for picking ducks.” This kind of engineer-specific romance is truly endearing.

Beyond that, the source code exposed many hidden features and clever undisclosed details. For instance, the internal codename for the project seems to be “Tengu,” and feature flags use gemstone-inspired names like “Cobalt Blue Frost.” The code also hides a voice system using Deepgram Nova 3 technology and reveals unreleased tools like kairos, an autonomous agent that monitors GitHub, and ultraplan, responsible for task planning on remote servers.

However, the code also reflects the reality and “technical debt” faced by all large-scale projects. A single main.tsx file reaching 800KB with over 4,600 lines of code, and as many as 460 “eslint-disable” comments, all point to the team’s compromises while rushing for delivery. They are still actively calling over 50 functions labeled “deprecated” in production, and the validation files contain 9 empty blocks that catch errors but do nothing. One member named Ollie even admitted in a comment that a certain piece of code “added a lot of complexity and it’s unclear if it actually improves performance.” This is actually quite normal—perfect code only exists in textbooks; what truly drives business are those efforts that are “just good enough to work.”

Ollama Fully Embraces Apple Silicon: A Local Computing Performance Leap

Beyond the intense competition in cloud computing, local development is equally significant. For those who value data privacy and offline operations, Ollama has been an indispensable tool. Recently, the official update for Apple Silicon support was released, fully integrating Apple’s proprietary machine learning framework, MLX.

How big is the impact of this change? Simply put, Mac users can now enjoy unprecedented execution speeds. Whether running personal assistants like OpenClaw or operating coding agents like Claude Code, OpenCode, or Codex, the smoothness has significantly improved.

Shifting heavy computing tasks directly to local hardware not only reduces reliance on internet connectivity but also makes daily development exceptionally smooth. This trend of seamless hardware and software integration is quietly changing the daily habits of every tech professional.

Q&A

We’ve compiled some key Questions and Answers (Q&A) to help readers quickly grasp these tech highlights:

Q1: What is the amount and valuation of OpenAI’s recent funding? What is their core product plan for the future? A1: OpenAI just announced a funding round of $122 billion, soaring its post-money valuation to $852 billion. With this capital, OpenAI is actively pushing to integrate ChatGPT, Codex (coding assistant), and web browsing into a powerful unified “Superapp.”

Q2: What competitive advantages does Google’s Veo 3.1 Lite offer in the video generation market? A2: The biggest advantage of Veo 3.1 Lite is its extreme cost-effectiveness. it maintains the same generation speed as the high-end Veo 3.1 Fast model, but at over 50% lower operational cost. Furthermore, it still supports 16:9 and 9:16 aspect ratios, as well as high-definition output at 720p and 1080p, precisely solving the pain point of high costs for developers generating videos.

Q3: The Claude Code source code was accidentally leaked. What interesting holiday easter egg was included? A3: Users found a “Tamagotchi” system called /buddy hidden in the terminal within the leaked source code. The code shows the salt for this system was set to “friend-2026-401,” confirming it was an April Fool’s event planned for April 1st. Users can hatch ASCII pets like ducks, capybaras, and dragons. Engineers even converted the word “duck” to hex code to evade internal code checks.

Q4: What real “technical debt” can be seen from the leaked Claude Code? A4: The source code reveals many “as long as it runs” development compromises. For example, the main.tsx file for handling messages is 800KB with over 4,600 lines of code, including 460 comments to force-disable linter warnings. Additionally, the team still calls over 50 “deprecated” functions in production, and an engineer even admitted in a comment that a certain piece of code “added complexity and it’s unclear if it actually improves performance.”

Q5: What major update did Ollama release for Mac devices? A5: Ollama has fully updated and integrated Apple’s proprietary MLX machine learning framework to perfectly support Apple Silicon. This change allows Mac users to enjoy significantly faster execution speeds and smoothness when performing local high-load tasks, such as running the OpenClaw assistant or operating coding agents like Claude Code, OpenCode, and Codex.