In partnership with
News of the day
1. Anthropic's Claude AI can now directly control your computer, handling tasks beyond standard app integrations for enhanced productivity. → Read more
2. Databricks launched Lakewatch, an open, agentic SIEM to unify security, IT, and business data. Lakewatch enables AI detection and response, reducing costs and eliminating vendor lock-in. The Open Security Lakehouse Ecosystem includes leading partners to enhance threat response. → Read more
3. Nvidia CEO Jensen Huang calls the idea of AI destroying software 'ridiculous,' stating AI agents will use existing software, not replace it. Nvidia is adapting its architecture. → Read more
4. Mirage raises $75m to expand its ai video editing tools, targeting creators and businesses with new models and a freemium model. → Read more
Our take
Hi Dotikers!
Last week, we talked about Channels, the Anthropic feature that turns Claude Code into a persistent agent, capable of receiving instructions from Telegram or Discord without a human sitting in front of their screen. The logic is the same today, but Anthropic is crossing a new threshold: Claude can now take direct control of your macOS desktop, open applications, navigate a browser, fill in spreadsheets, and execute tasks you would normally do yourself, in your place.
The feature is available as a research preview in Claude Cowork and Claude Code. It relies in part on technology from Vercept AI, a startup specializing in computer interface control that Anthropic acquired a little under a month ago. Kiana Ehsani, co-founder of Vercept, shipped a first product in less than four weeks after the acquisition, which says a lot about the pace at which Anthropic intends to move.
The design principle is interesting: Claude only takes control of the desktop as a last resort, when no native integration (Slack, calendar, connected tools) is available. It's an elegant way to frame the thing, but it shouldn't obscure what this concretely implies: a language model that sees your screen, clicks on your behalf, and acts within your real digital environment.
OpenAI had attempted something comparable with its ChatGPT agent, limited to the browser, and the results hadn't lived up to the ambitions. Anthropic is aiming much broader with full desktop control, which proportionally increases the attack surface. And this isn't a theoretical concern: just days after the initial launch of Cowork, security researchers had already documented a prompt injection enabling the exfiltration of confidential files. The agent working for you can also, without knowing it, work against you.
The real question isn't whether Claude can do this technically, it's impressive enough to answer yes. The question is whether we are collectively ready to hand over the keys to our work environment to a system that remains, by definition, probabilistic. Fixing up your calendar is one thing. Accessing your files while your back is turned is an entirely different conversation.
Alexis
The ops hire that onboards in 30 seconds.
Viktor is an AI coworker that lives in Slack, right where your team already works.
Message Viktor like a teammate: "pull last quarter's revenue by channel," or "build a dashboard for our board meeting."
Viktor connects to your tools, does the work, and delivers the actual report, spreadsheet, or dashboard. Not a summary. The real thing.
There’s no new software to adopt and no one to train.
Most teams start with one task. Within a week, Viktor is handling half of their ops.
Meme of the day
