Good Morning! Here's what's happening:
Okara launched the world's first influencer agent, finding creators, managing outreach, running campaigns, and handling payments automatically
ChatGPT for iOS updated Codex Mobile with a workspace file browser, MCP approval controls, and LaTeX rendering
NVIDIA published SpatialClaw, a training-free spatial reasoning agent that outperforms prior SOTA by 11.2 points across 20 benchmarks
Plus: How to connect Claude to Gmail, Drive, Slack, and more
4 new AI tools worth trying today
AI MODELS
Okara introduced the world's first influencer agent, an AI that handles creator discovery, outreach, campaign management, payments, and content publishing from start to finish.
Tell the agent what you want to promote and it handles the rest. It finds relevant creators that match your brief, sends outreach on your behalf, and negotiates terms without you writing a single email.
7 specialized sub-agents run your marketing across 6 channels every 24 hours simultaneously, covering Reddit, X, LinkedIn, SEO, Hacker News, UGC Videos, and a dedicated Influencer Agent managing the full creator pipeline.
Pricing starts at $98/month, compared to $160K/year for a human marketing team by their own breakdown. The agent runs campaigns, processes payments, and tracks results without any additional headcount.
Available now at okara.ai/influencer. Brief your first campaign, define your brand assets and requirements, and get content live without speaking to a single human.
Influencer marketing has always required either a big budget or weeks of manual work, finding creators, writing pitches, negotiating rates, managing deliverables, and processing payments separately. An agent that runs the entire pipeline end-to-end removes every one of those steps. For solo creators and small teams trying to grow, this is the kind of automation that used to be reserved for brands with a full marketing department behind them.
AI MEMORY
OpenAI shipped a major Codex Mobile update for ChatGPT on iOS, adding a workspace file browser, MCP approval controls, and LaTeX rendering inside the app.
A new workspace file browser lets you preview files and link workspace paths directly into your prompts. No more switching to Finder, copying a path, and pasting it manually into a new chat.
MCP approval controls now let you approve or deny agent-requested actions per chat or globally across all chats, critical for anyone running Codex autonomously who wants to stay in control of what it actually executes.
LaTeX now renders properly inside Codex messages and plans, so equations and technical notation display as formatted math instead of raw text cluttering your output.
A new directory picker lets you choose your workspace folder at the start of each thread, so Codex loads the right project context automatically instead of reading from whatever was last open.
This update closes a meaningful gap between what Codex does on desktop and what it can do from your phone. The MCP approval controls are the most critical addition. As coding agents take more autonomous actions, having granular control over what they're actually allowed to execute becomes essential.
AI AGENTS
NVIDIA Research published SpatialClaw, a training-free agent that writes Python in a persistent kernel to handle complex visual and spatial reasoning tasks.
Instead of calling a fixed set of pre-defined tools, SpatialClaw writes Python inside a persistent kernel. It composes perception modules on the fly, inspects intermediate results, and revises its strategy mid-task without starting over.
Perception outputs become ordinary Python variables the agent can reuse and combine with libraries like NumPy and SciPy, giving it far more flexibility than standard tool-calling agents where every output gets discarded after use.
With no benchmark-specific or model-specific tuning, it outperforms the previous leading agent by 11.2 points across 20 visual benchmarks, a significant margin without any fine-tuning or domain adaptation.
Results hold consistently across 6 different model backbones, meaning the performance gain comes from the architecture itself and not from the underlying model, making it broadly applicable.
Most spatial reasoning agents are brittle because they rely on a fixed toolkit. If a task falls outside it, performance collapses. SpatialClaw treats code as the interface instead, which means it adapts its strategy dynamically based on what it's actually seeing. The benchmark results without any fine-tuning are the most telling part, 11.2 points ahead of the previous best agent, across 20 benchmarks, on 6 different model backbones.
PREMIUM GUIDE
🔌 Claude Connectors: How to stop switching tabs and let Claude read your apps directly
How to connect Claude to Gmail, Drive, Slack, and more.
I spent 45 minutes preparing for a 20-minute call last week.
Not preparing what to say. Preparing the information. Opening Gmail for the thread history. Switching to Drive for the shared document. Checking Calendar for the last time we met. Building the brief manually from three different apps before I could even start the actual work.
All of that work just to give Claude information it could have pulled itself, if I'd had connectors set up.
I set them up that week. I haven't gone back.
A connector is a one-click bridge between Claude and an app you already use. Gmail, Google Drive, Google Calendar, Slack, Notion, Gamma, and 200+ more. When a connector is on, Claude reads that tool directly. You stop copy-pasting. You stop switching tabs. You just ask.
How to connect your first app.
Four steps. Under two minutes.
Open Claude. Click the + inside the chat input bar.
Click Connectors → Add Connector.
Search for your app. Click it. Log in.
Open a new chat. Click + in the chat bar → Connectors → toggle your app on.
One setting worth changing right away: go into each connector's settings and set it to "Always Allow." Otherwise Claude asks permission every single time it uses it. Gets old fast.
Want to read the full guide?
What comes after is the actual guide: the 6 connectors worth turning on first, the exact prompts for each one, and the three rules that keep them from working against you.
✓ Full archive of premium guides with ready-to-use prompts
✓ Structured AI courses (step-by-step, start-to-finish)
✓ Every upcoming premium tutorial
Slack rolled out native charts in Slackbot - prompt it to generate a chart from your data and it builds one directly in the conversation, no extra tools or tab switching needed.
Manus added a message queue feature - prompts sent while a task is running execute automatically once the current task finishes.
Framer launched version 3.0 with built-in AI agents, branching workflows, a native community hub, and a completely rebuilt interface.
Okara: The first influencer agent - finds creators, handles outreach, manages campaigns, and processes payments end-to-end.
ChatGPT for iOS: Codex Mobile updated with a workspace file browser, MCP approval controls, and LaTeX rendering.
Manus: Autonomous AI agent that now lets you queue prompts while a task runs so nothing gets lost mid-session.
Framer: Web design platform that launched 3.0 with built-in AI agents, branching workflows, and a completely new interface.
THAT’S IT FOR TODAY
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