Google launched Gemini 4 Argon, a model that writes up to 1 million tokens in a single response and beats GPT-6 Astra on key benchmarks, OpenAI launched ChatGPT Dots, always-on agents with their own cloud computer that keep working while you're offline, and Free Claude Code lets you run Claude Code through free providers or your own machine at zero cost. And that's not all, here's everything in AI this week:
Orca is a free desktop app for running multiple AI coding agents at the same time. You can run Codex, Claude Code, OpenCode, and Pi side by side, each in its own git worktree, and manage all of them from your phone.
Parallel Agents: Run several coding agents at once, each in an isolated worktree so they never edit the same branch.
Phone Control: A mobile companion app notifies you when an agent finishes and lets you send follow-ups from anywhere.
Bring Your Own Subscription: No login required. It works with 30+ CLI agents using the subscriptions you already pay for.
Try it now → https://www.onorca.dev/download
Google has launched Gemini 4 Argon, the first model in its Gemini 4 generation and its most powerful model yet. It can write up to 1 million tokens in a single response and beats GPT-6 Astra and Claude Opus 5.5 on most of Google's published benchmarks.
1 Million Token Output: Up from 64,000 on earlier Gemini models. GPT-6 Astra and Claude Opus 5.5 cap a single response at 128,000 tokens.
Benchmark Leader: Scored 77.9% on the DeepSWE v1.1 coding benchmark, ahead of Opus 5.5 (74.2%) and GPT-6 Astra (74.1%), and led or tied on 12 of 18 published tests.
Cheaper Than Astra: Launch pricing is $2 per million input tokens and $10 per million output, about one fifth of GPT-6 Astra's price.
Learn more → https://deepmind.google/models/gemini/
OpenAI has launched Dots, always-on agents that live inside ChatGPT. Each one runs on GPT-6 Astra with its own cloud computer and keeps working toward your goals 24/7, even when you're offline.
Its Own Computer: Each Dot gets a dedicated cloud computer and browser. Access to your own computer is optional and off by default.
4,000+ Apps: Dots connect to more than 4,000 apps through ChatGPT plugins, and you can message them in ChatGPT, Slack, or Teams.
Works While You're Away: Dots keep researching and working between conversations, and chatting with one does not count toward ChatGPT usage limits.
Try it now → https://chatgpt.com/features/dots
Free Claude Code is an open-source proxy that sends your Claude Code traffic to free providers like DeepSeek and Kimi, or to models running on your own machine. You keep the same terminal and the same VS Code extension, at zero cost.
Drop-In Proxy: Claude Code keeps working exactly as before while the proxy forwards its requests to the provider you choose.
Free and Local Options: Supports NVIDIA NIM's free tier, OpenRouter free models, DeepSeek, Kimi, and fully local models through Ollama, LM Studio, or llama.cpp.
Per-Model Routing: Send Opus, Sonnet, and Haiku requests to different models, and pick them from Claude Code's built-in model picker.
Try it now → https://github.com/Alishahryar1/free-claude-code
AutoHarness is a free plugin that gives Claude Code self-learning skills. It watches your real sessions, writes skills from what you do, and maintains them without any manual work.
Learns From Your Sessions: It turns patterns from your actual work into skills and saves them to your project or global skills folder.
Merges Duplicates: When two skills cover the same scenario, it combines them instead of keeping near-copies.
Deletes Unused Skills: Skills that stop getting used are removed, and it never touches skills you wrote yourself.
Try it now → https://github.com/tigerless-labs/autoharness
Xiaomi has open-sourced MiMo-V2.6-Pro, the highest-ranked open-weight model in the world. It has 1 trillion parameters and is completely free under the MIT license.
Massive but Efficient: A Mixture-of-Experts model with over 1 trillion total parameters, only about 42 billion active per token, and a 1 million token context window.
Number 1 Open Model: Scored 46.32 on the Artificial Analysis Intelligence Index, the highest of any open-weight model at launch, ahead of Kimi K3 and GLM-5.3.
Free and Multimodal: MIT-licensed with native text, image, video, and audio input, plus more than 7,000 reinforcement-learning environments released alongside it.
Try it now → https://mimo.mi.com/
Thanks for making it to the end! I put my heart into every email I send. I hope you are enjoying it. Let me know your thoughts so I can make the next one even better.
See you tomorrow :)
Dr. Alvaro Cintas
✓ Full archive of premium guides with ready-to-use prompts
✓ Structured AI courses (step-by-step, start-to-finish)
✓ Every upcoming premium tutorial
✓ $500+ worth of AI tools