A former Microsoft Research leader has written that the AI Frontiers (AIF) sub-organization, a roughly 50-person group formed in 2023 after ChatGPT's release to pursue coordinated AI research, has ended. The group had worked on agent frameworks like Autogen and the Fara computer-use agent, alongside fundamental research such as the Next Latent objective, Dion3, and the Full Bandwidth Transformer, and had built a shared monorepo used across Microsoft Research.
hunch.net
· 2026-11-07
OpenAI has released GPT-6.1 Terra, a smaller model built for computer-use tasks like form-filling, video editing and reading technical drawings. The company says it runs over three times faster than GPT-6 Astra and GPT-6.1 Sol, while API pricing drops 50% to $1 per million input tokens and $5 per million output tokens. OpenAI reports Terra's error rate fell from 53.6% to 16.7% at low reasoning effort compared to GPT-5.6 Terra, and its coding score on Terminal-Bench 4.0 more than doubled to 50.3%.
bleepingcomputer.com
· 2026-10-26
Startup Hark, founded less than a year ago by Brett Adcock, released Hark Pro to the public today with a free tier and a paid subscription for heavier use. The app presents a full-screen interface with a central chat window, a feed of suggested actions, and visual 'panels' showing dashboards for things users care about, and it can connect to email, calendars, hard drives and payment accounts to carry out digital tasks on a user's behalf.
techcrunch.com
· 2026-10-06
Google is reportedly rolling out a new Tasks mode within the Gemini desktop app, according to TestingCatalog, which separates regular chat conversations from more autonomous, multi-step tasks. The update includes new Computer Use settings allowing users to specify which desktop apps Gemini can and cannot control, plus a more visible option to launch projects from local folders. A Gemini Live button has also been spotted inside Tasks, suggesting users may be able to talk to Gemini while it operates across the computer.
androidauthority.com
· 2026-09-24
Cua released CUA-S1, a specialized model for computer-use decisions, alongside Cua Driver, a toolset letting AI agents inspect and control native desktop apps and browsers on macOS, Windows, and Linux via CLI, MCP, or SDKs. The company also offers Cua Fleets, cloud-hosted sandbox desktops provisioned through run.cua.ai, so developers can test agents on isolated Linux environments without local hardware.
github.com
· 2026-09-19
A new open-source tool called typesafe-computer-use lets a Mac execute plain-English goals by reading the screen state directly rather than sending screenshots to a large multimodal model. It uses a lightweight classifier from TypeSafe to pick the next action from a short list of options, only invoking a text-generation model when free-form input is actually required. The developers report costs near 1/50th of a cent per decision step, versus tens of cents per step for a comparable Claude Opus-based screenshot approach.
github.com
· 2026-09-16
GoKawiil