In partnership with

Welcome back!

OpenAI just unveiled ChatGPT Agent, a powerful new tool that can browse sites visually, code, analyze, and automate workflows. It gives ChatGPT its own digital workspace, complete with API access, and it uses it to complete multi-step tasks. It’s the next stage in the evolution of Operator and Deep Research. Let’s unpack…

Today’s Summary:

• 🚀 OpenAI launches ChatGPT Agent

• 🧑‍💻 Kimi K2 open-source AI

• 🗣 Mistral AI upgrades Le Chat

• 🧑‍🎨 OpenAI adds high image fidelity

• 📈 Claude launches financial services AI platform

• 📞 Google adds AI to call businesses

• 🛠️ 2 new tools

TOP STORY

OpenAI launches ChatGPT Agent

The Summary: OpenAI has released ChatGPT Agent, a “cutting-edge and experimental” evolution of Operator and Deep Research that allows ChatGPT to act autonomously across tools like GUI browsers, terminals, and APIs. It can carry out complex workflows like research, coding, shopping, or preparing slide decks within its own internal environment.

Key details:

• ChatGPT Agent runs its own internal “computer” with a GUI web browser connected to apps like Gmail

• Spreadsheets and slides made for you: users can request reports, get them as editable decks or .xlsx files, and then tweak the final 2%

• One-click scheduled automations: users can create tasks like reports, email digests, or competitor scans and schedule them to run every Monday

• Prompt injection risk is real: invisible prompts on websites could trick the agent into sharing private info, like leaking Gmail data. OpenAI has added safeguards like “Watch Mode” and specific training. Use with caution.

• Operator is being phased out. Deep Research is folded into Agent

• Rolling out now to Pro, Plus, and Team. Not yet available in the EU and Switzerland. Enterprise and Education coming soon.

Why it matters: ChatGPT continues to move toward a general agent executor. This turns AI from a source of information into a source of outputs, doing online work with minimal friction. The efficiency gains sound great, but we should consider that “98% right” could mean it’s just as difficult to spot the 2% remaining errors. The future may be about which agent we trust more to act without errors.

FAQ

FROM OUR PARTNERS

The Missing Tool in Your AI Stack

Create How-to Videos in Seconds with AI

Stop wasting time on repetitive explanations. Guidde’s AI creates stunning video guides in seconds—11x faster.

• Turn boring docs into visual masterpieces

• Save hours with AI-powered automation

• Share or embed your guide anywhere

How it works: Click capture on the browser extension, and Guidde auto-generates step-by-step video guides with visuals, voiceover, and a call to action.

Best part? It’s 100% free.

👉 Download the extension now

OPEN SOURCE

Kimi K2 brings open source agentic intelligence

The Summary: Moonshot AI launched Kimi K2, the largest state-of-the-art open-weight model to date. With a trillion parameters, it matches or beats top models like Claude Sonnet 4 on coding and math benchmarks. Now available to developers and researchers, it could mark a second “DeepSeek” moment for open source.

Key details:

• Kimi K2 is a 1T-parameter Mixture-of-Experts model with 32B active per inference, trained with the new MuonClip optimizer

• Scores 65.8% on SWE-bench Verified benchmark (code bug fixing), beating GPT-4.1 and just behind Claude 4 Sonnet

• Very hardware-intensive to self-host (958GB), but can run with 4-bit quantization on two M3 Ultra Mac Studios

• Available for self-hosting via GitHub

Why it matters: Kimi K2 is the largest open-weight model crossing the threshold of enterprise capability for non-reasoning models. Its performance shows that massive reasoning chains aren’t the only path to innovation.

MISTRAL AI

Mistral leans in with voice, research, and code

The Summary: Mistral has rolled out major upgrades to its chatbot Le Chat, bundling deep research mode, multilingual logic, image editing, and an open-source voice model called Voxtral. While many features echo those of ChatGPT or Gemini, Mistral’s privacy-first approach and on-prem enterprise compatibility make it a rare breed. With these additions, Le Chat now functions more like a full-stack AI workspace than just another chatbot.

Key details:

• Deep Research mode builds structured, source-backed reports from web queries

• Voxtral, the open-source voice model, transcribes speech with better performance per dollar than OpenAI Whisper

• Prompt-based image editing with strong visual consistency

• Le Chat Projects offer long-term workspace contexts with memory, tool presets, and file upload, like NotebookLM

• Devstral Medium, Mistral’s new coding model, beats Gemini 2.5 Pro on SWE-Bench Verified and runs cheaper

• All features are live today on web, iOS, and Android

Why it matters: Mistral is building a complete AI ecosystem optimized for sovereignty, cost-efficiency, and developer freedom. Le Chat now offers structured research, reliable voice mode, and image editing in one interface, with no cloud vendor lock-in.

QUICK NEWS

Quick news

• OpenAI rolls “High Input Fidelity” in the API for precise image editing

• Anthropic launches Claude for Financial Services platform

• Google Search adds AI-powered business calling

TOOLS

🥇 New tools

• Finlens - AI accounting for founders and accountants

• Mozart - AI-assisted music production