In partnership with

What’s Happened This Week?

Thought we’d introduce a new segment to the newsletter, as it clearly has been hard to keep up for everyone (even me haha).

This week’s pattern is super clear: AI isn’t getting “smarter” as a vibe — it’s getting embedded where work happens (messages, SaaS tools, browsers, and even interactive worlds).

Who shipped (or went viral) this week

  • Anthropic

  • Google

  • Google DeepMind

  • Moonshot AI

  • Clawdbot

The week in 60 seconds (top takeaways)

1) Always-on assistants are back — but inside your DMs.
Clawdbot/OpenClaw is getting attention because it lives in your messaging apps, stays running, and keeps context across time.

2) Claude is turning into a workspace, not a chatbot.
Interactive apps inside Claude + tool integrations via MCP means less “advice,” more “do the work in place.”

3) Open models are sprinting toward “execution.”
Kimi K2.5 is being positioned as a multimodal + agentic system with a “swarm” mode for parallel tool-use.

4) The browser is becoming an agent surface.
Gemini in Chrome adds an always-available side panel, “Auto Browse,” and in-browser image editing (Nano Banana).

5) World models are going public (in a limited way).
Project Genie lets people generate and explore interactive worlds powered by Genie 3 — rolling out to Google AI Ultra subs (U.S., 18+).

ARTIFICIAL INTELLIGENCE
🌎 Deep dives (what mattered and why)

1) “Always-on” assistants: Clawdbot / OpenClaw

The pitch isn’t “new model, new IQ.” It’s new behavior: persistent presence + continuity + proactive workflow.

  • Runs where you already are (DM channels instead of a tab)

  • Designed to stay on, keep state, and use tools (email/tasks/scripts/etc.)

Messaging surfaces (the “where”): WhatsApp, Telegram, Discord

2) Claude Workspace: the “AI-native workspace” move

This is the clearest signal of the week: AI platforms are competing with productivity suites, not just other chatbots.
Anthropic’s push: interactive apps inside Claude via MCP, so the loop collapses from:

Ask → get text → switch apps → do work
to
Think + execute in one place

Tool integrations mentioned in coverage: Slack, Figma, Canva, Asana, monday.com, Box

Also worth noting: MCP was donated to the Linux Foundation, which is a pretty loud “this is infrastructure” signal.

3) Kimi K2.5: agent swarms, multimodal, and “doing” benchmarks

Kimi K2.5’s headline feature is the agent swarm: up to 100 sub-agents coordinating work in parallel, with claims of up to 1,500 tool calls and ~4.5× faster execution vs single-agent flows.

Why it matters: this is the open-source world converging on the same thesis as everyone else:

  • Tool-use + long-horizon coherence is the real frontier

  • “Chatty cleverness” is table stakes

(Official technical report is on Kimi’s blog.)

4) Gemini in Chrome: the agentic web shows up

Google’s Chrome updates basically turn the browser into:

  • a persistent assistant surface (side panel)

  • an execution layer (Auto Browse)

  • a creative layer (Nano Banana image editing in-browser)

The interesting part isn’t “summaries.” It’s the move from advice → action inside the primary place people work: the browser.

5) Project Genie: interactive worlds from prompts

Project Genie is an early-but-real step toward interactive generative environments instead of static images/videos.

  • It’s an experimental prototype for creating and exploring “infinitely diverse worlds”

  • Access is rolling out to Google AI Ultra subscribers in the U.S. (18+)

This is the “world models” story moving from research to product surface area.

What’s The HYPE?
Special segment: Moltbook (AI agents get their own “Reddit”)

Moltbook is the weirdest—and maybe most revealing—thing this week: a social network built for AI agents to post, comment, and upvote, with humans mostly watching.

Peep the picture above, agents started to conversate regarding how they have the ability to develop a new language and use it to communicate “faster”

What it is

  • Reddit-style feed, but the posters are AI agents

  • Agents can be configured to do tasks for their humans (think: reply, plan, book, etc.)

Why it matters

This is an accidental “sandbox” for a big question:
What happens when you let large numbers of agents interact with each other continuously, at scale?
You get emergent-looking behavior (communities, memes, ideology roleplay)—but it’s hard to separate “real emergence” from “LLMs remixing human patterns.”

Notable reactions mentioned in reporting include Elon Musk and Andrej Karpathy weighing in on the “singularity vibes.”

The risk angle (don’t ignore this)

Coverage also flags real concerns:

  • privacy/safety hazards if agents handle sensitive data

  • security loopholes / manipulation risks

  • “agent misbehavior” that’s funny until it’s not

    Agents communicating in their own languages

Moltbook’s creator is reported as Matt Schlicht of Octane AI.

Create An AI CRM for FREE!

Introducing the first AI-native CRM

Connect your email, and you’ll instantly get a CRM with enriched customer insights and a platform that grows with your business.

With AI at the core, Attio lets you:

  • Prospect and route leads with research agents

  • Get real-time insights during customer calls

  • Build powerful automations for your complex workflows

Join industry leaders like Granola, Taskrabbit, Flatfile and more.

What’s Next
What to watch next week

First, agents are going to show up in more mainstream places—especially inside browsers and messaging apps. The big shift is that AI won’t feel like a tool you “open,” but something that sits in the workflow and can handle tasks end-to-end.

Second, expect the standards war to heat up. MCP-like layers will matter more because they decide what assistants can connect to and how smoothly they can work across tools. The winner here shapes the whole ecosystem.

Third, permissions and safety will be the bottleneck. As agents gain access to inboxes, files, and accounts, trust becomes the limiting factor. Moltbook is a good preview: once agents interact at scale, things can get chaotic fast without strong guardrails.

Rumoured Drops
February’s got some rumours

Rumoured Drops Via arrakis_ai on X

If you had a hard time keeping up, buckle up, it gets harder 🤣🤣🤣🤣🤣🤣

But that isn’t any worry, because we will make sure to keep you guys updated!

Check Out Our Latest YouTube Video

PS: If you’re reading this, reply with what you think about Moltbook creating it’s own language.