This week AI stopped answering questions and started doing the work. Agents moved into your apps and even your phone, while the sector’s first real price war kicked off.
I’ve filtered out the noise for you
Five stories, no filler and no one’s hot takes. Here is what actually mattered in AI this week.
1. ChatGPT can now do your job while you sleep
On July 9, OpenAI launched ChatGPT Work, an agent that takes actions inside your apps and files and can work on a project for hours without supervision. It arrived alongside the new GPT-5.6 model family (Sol, Terra, and Luna) and goes head to head with Claude Cowork. The chatbot was the demo; now the agent is the one doing the work.
2. Was Meta’s free AI just bait?
After years of free and open models, Meta started charging for access to its AI with Muse Spark 1.1, focused on code and agents. The launch includes 20 dollars in credits and prices below rivals like Grok. The era of open source as a public relations strategy looks like it is over.
3. China built the first phone that runs itself
StepFun and ZTE’s Nubia brand are getting ready to unveil the first phones with an AI agent: devices whose AI operates the apps for you instead of just answering questions. StepFun reveals its phone on July 13 and others debut at the WAIC in Shanghai, July 17 to 20. Each one bills itself as the world’s first AI agent smartphone.
For sales teams tired of cold leads, slow customer responses, and manual processes, Dapta is the ultimate tool.
Dapta is the leading platform for creating AI sales agents specifically designed to increase inbound lead conversion. Respond to your leads in less than a minute with voice AI and WhatsApp that converts.
If you want your team to sell more while AI handles the complex stuff, you have to try it.
4. OpenAI just killed Atlas before its first birthday
OpenAI announced it will shut down its ChatGPT Atlas browser on August 9, less than a year after launch, and fold its agent browsing capabilities into its new desktop super app. CEO Fidji Simo is cutting side projects to concentrate everything on ChatGPT. The company frames it as a lesson, not a failure.
5. Google just made making AI images dirt cheap
On June 30, Google launched two models: Nano Banana 2 Lite, its fastest and cheapest image model (around 4 seconds per image at about 0.034 dollars), and Gemini Omni Flash for generating and editing video with conversational commands. Both aim to cut the cost per creative operation and win on volume.
Bonus: Content of the Week
Bloomberg released a mini documentary about Anthropic featuring CEO Dario Amodei and his sister Daniela. The video digs into big questions about the future of work and AI.
