Tech • AI • Robotics • Game
OpenAI’s announcements (including Dots!) and Anthropic’s: It’s heating up!
9.4/10TL;DR
OpenAI has unveiled a broad shift toward cloud-based AI agents, collaborative workspaces and platform integrations, while Anthropic and AMD made parallel moves that intensify competition across models, infrastructure and capital markets.
KEY POINTS
OpenAI pushes autonomous agents
OpenAI’s headline move is a new class of autonomous agents called Dots, designed to act with limited human prompting and trigger tasks on their own. The shift marks a move away from AI as a simple assistant toward AI as a co-worker embedded in daily workflows, with major implications for office work, customer handling, coding and task automation.
From local tools to cloud execution
The broader strategy is to move more of the work stack into OpenAI’s cloud. Codex now runs in the cloud rather than primarily alongside a user’s local setup, reflecting a model in which inference, memory, customization and task execution increasingly sit on the provider’s infrastructure rather than on company devices.
Decision API targets fast machine choices
OpenAI also introduced a Decision API built for rapid, bounded outputs instead of conversational replies. Using the lightweight Luna model, it is aimed at classification and predefined choices, such as selecting an answer or returning a confidence score, a sign that vendors are productizing AI for routine operational decisions.
New model pricing reshapes usage
A new model, GPT-6.1 Sol, is positioned as far cheaper than Astra while approaching it on many practical tasks. Pricing cited for output was $10 per million tokens for Sol versus $50 for Astra, suggesting that near-premium capability is becoming dramatically less expensive for businesses deploying AI at scale.
ChatGPT becomes a workspace
With ChatGPT Spaces, OpenAI is building a collaborative environment where humans and AI agents can work on shared pages, files and presentations. The product vision is an office suite where teams include both people and AI, extending OpenAI beyond chatbot use into project execution and team coordination.
Integrations aim at platform control
OpenAI is expanding direct integrations into workplace tools such as Slack and Teams, while also enabling plugins and account-based connections to third-party services. That would let users access partner tools through ChatGPT, potentially using their OpenAI subscription, a platform strategy that could pull design, productivity and business software into one interface.
Higher-tier subscriptions reflect enterprise demand
The company is adjusting paid plans to match heavier enterprise usage. A $200 tier returns with lower quotas than before but compensation credits for some existing users, while a new $500-per-month plan targets customers running large volumes of AI work, especially those managing multiple agents and collaborative spaces.
Europe remains behind on availability
Many of the new features are not yet available in Europe, though the new models are accessible. The staggered rollout underscores how regulatory and product launch timing continue to leave European users behind US releases for some advanced AI workplace tools.
Anthropic answers with Claude Sonnet 5.5
Anthropic released Claude Sonnet 5.5, a lower-cost model that appears close to Opus 5 on many tasks and stronger than its predecessor in agentic coding. In one cited independent ranking from Artificial Analysis, GPT-6.1 Sol scored 52, Opus 5 scored 58, and Sonnet 5.5 landed above them, reinforcing Anthropic’s position in high-quality model performance.
AMD buys World Labs in a bet on embodied AI
AMD has acquired World Labs, the AI startup founded by Fei-Fei Li, in a deal described at over $8 billion despite a staff of roughly 60 to 70 people. The rationale is talent and long-term research: world models and spatial intelligence could become essential for robotics and next-generation AI systems, giving AMD a potential shortcut in its rivalry with Nvidia.
Anthropic files for an IPO amid surging revenue
Anthropic has filed for an initial public offering after reporting rapid growth, including $4.6 billion in annual revenue previously and $11.5 billion in revenue in Q2 2026 alone. The filing reportedly devotes about 80 pages of a 260-page document to AI risks such as cyber abuse, manipulation and loss-of-control scenarios, highlighting the tension between explosive commercial growth and the industry’s own warnings.
CONCLUSION
The latest moves show the AI market shifting from standalone models to full operating platforms built around agents, subscriptions and cloud control. That raises the commercial stakes for OpenAI, Anthropic and AMD, while intensifying questions about jobs, market power and who controls the infrastructure of future work.
Ask a question