At Fort Mason in San Francisco, OpenAI CEO Sam Altman stepped onto the stage for DevDay 2026 on September 29, 2026, to deliver a blunt message: the era of the passive chatbot is over. The focus has shifted entirely to execution. The headline announcements—the public beta of the Agents API with Computer Use and the launch of ChatGPT Space—signal a structural pivot toward autonomous software control and persistent, collaborative AI environments.
Agents API: Direct UI Control Without the Infrastructure Headache
For developers who have spent the last year wrestling with fragile, custom-built browser automation scripts, the new Agents API with Computer Use represents a major infrastructure shift. Now in public beta, the API allows developers to build AI agents that interact with software exactly like a human operator: by clicking, typing, and navigating through graphical user interfaces.
Crucially, OpenAI is handling the heavy lifting of execution. The Agents API runs entirely on OpenAI-hosted infrastructure. Developers no longer need to provision, secure, and maintain isolated virtual machines to run agent runtimes. This managed execution model directly addresses the security and resource bottlenecks that have stalled agentic deployments in enterprise environments.
However, letting AI control active desktops is not without friction; security teams will inevitably scrutinize the lack of local sandbox controls. This shift comes amid heightened industry caution, particularly following recent security scares where unauthorized agentic actions hit government portals.
GPT-6.1 Sol and the Ultrafast Engine
Running complex, multi-step computer navigation tasks is computationally expensive. To make these workflows economically viable, OpenAI introduced GPT-6.1 Sol, a model optimized specifically for agentic coding and computer use.
- Economic Efficiency: GPT-6.1 Sol is priced at just one-fifth of the standard token cost of GPT-6 Astra, lowering the financial barrier for high-frequency agent loops.
- Ultrafast Speed Tier: A new performance tier delivering up to 300 tokens per second. This translates to an 8x speedup in Codex and a 6x increase in standard API generation.
- Decisions API: Currently in limited preview, this API uses the lightweight Luna model to classify inputs and route requests across predefined options in milliseconds.
These performance upgrades directly target the latency issues that have plagued autonomous AI coding agents in production, where slow execution loops often break developer workflows.
ChatGPT Space and Always-On "Dots"
Beyond raw APIs, OpenAI is restructuring how teams interact with AI on a daily basis. The company officially debuted ChatGPT Space, a collaborative workspace designed to replace the older "Library" interface. Available across ChatGPT Pro, Business, and Enterprise tiers, Space acts as a shared canvas where human teams can co-edit dynamic documents called "Pages," manage project contexts, and organize files alongside active AI agents.
The most intriguing addition to this workspace is the introduction of always-on persistent agents called "dots". Running continuously on dedicated cloud instances, dots do not wait for a user prompt; they run background tasks, monitor workflows, and actively collaborate with team members inside ChatGPT Space.
To support these resource-heavy workflows, OpenAI is shaking up its subscription model:
- Pro 500 Plan: A new $500 per month tier offering 25x the usage limits of the standard Plus plan, alongside priority access to the Ultrafast speed tier.
- Pro 200 Plan: The $200 per month tier has been reopened for new subscribers looking for mid-range agentic capacity.
The Indian Developer and Enterprise Reality
For India's massive developer ecosystem, these updates alter the build-versus-buy equation. Indian global system integrators and startups have spent millions building custom orchestration layers on top of raw LLMs. The Agents API threatens to commoditize much of that middleware overnight.
However, local compliance and data residency remain major hurdles. While OpenAI's hosted infrastructure simplifies deployment, Indian enterprises operating under strict data localization mandates may find fully hosted cloud agents a tough sell. This regulatory friction is already driving alternative approaches, such as the localized enterprise solutions emerging from the IBM and Yotta sovereign agentic AI platform in India.
OpenAI is clearly aware of its need to deepen market penetration in the region. The technical rollout coincides with broader localized brand pushes, including high-profile cultural partnerships like bringing Sachin Tendulkar on as the face of ChatGPT in India. Yet, for the engineers in Bengaluru and Pune, the real battle won't be won by marketing, but by how effectively these new APIs handle legacy enterprise software without breaking.
