Latest AI News: Agents Take Over (Nov 08)

Perplexity launched always-on AI agents called Teammates, Cursor started deploying full web apps from a text prompt, and the U.S. Department of Justice formed a Super Intelligence Task Force — all in one week. Here is the AI news that matters for practitioners this week.

Perplexity Teammates: Always-On AI Agents for $200/Month

Perplexity shipped persistent AI agents that own ongoing work instead of answering single prompts. Teammates launched on November 5, 2026, for Max subscribers on the web at $200 per month. These agents remember project context, carry assigned roles, work proactively without being prompted, and can message coworkers directly in Slack.

What makes Teammates notable is the shift from reactive to persistent. Instead of asking a question and getting an answer, you assign a Teammate a standing responsibility — sales research, customer support monitoring, operations tracking — and it keeps working. The agents run on top of Perplexity Computer and reuse existing Projects, Skills, Memory, and scheduled automations. They can also hand off tasks to other Teammates.

This is a direct move into the enterprise agentic workflows space that OpenAI and Anthropic already occupy. The pricing positions it against Microsoft Copilot's seat-based model and Anthropic's developer-first Claude Code. For practitioners evaluating agent orchestration tools, Teammates is worth testing if you already use Perplexity's ecosystem.

Cursor Ships Build Live Apps — Prompt to Hosted Web App

Cursor released Build Live Apps on November 5, 2026, and it changes what the IDE can do. You describe the app you want to a cloud agent, and it builds the application code, spins up a Postgres database, sets up authentication behind Cursor SSO, provisions hosting, and hands you a shareable live link.

The feature is available on all paid plans with cloud agents enabled. According to Cursor's changelog, the workflow goes: kick off a cloud agent, describe what you want, review the security checks, then publish. Apps can deploy from new or existing repos on Cursor Origin and GitHub. Neon provides the Postgres infrastructure, and deployment runs on Vercel.

This closes a real gap. As Cursor put it, agents made writing code cheap, but most apps never made it off localhost. Build Live Apps automates the deployment half. For anyone exploring AI code generation, this is a practical step toward full-stack agent-driven development — though hosting limits, traffic quotas, and portability remain unpublished.

OpenAI Images 2.6: GIFs, Faster Generation, New API Models

OpenAI rolled out Images 2.6 on November 4 across ChatGPT and Codex on all subscription tiers. The headline features: native animated GIF creation from prompts or images, persistent reference images (save photos of yourself, pets, or characters for reuse across chats), and 50% faster generation.

For developers, two new API models launched: GPT-Image-2.6 Flare (optimized for speed, generating images in as little as six seconds) and GPT-Image-2.6 Sunburst (optimized for quality, with better edit consistency and SVG export). Pricing stays the same as Images 2.5.

The scale number worth noting: OpenAI now generates over 4 billion images per week across ChatGPT, Codex, and APIs — up from 3 billion just eight weeks ago. On the Artificial Analysis image editing leaderboard, GPT Image 2.6 Sunburst leads with an Elo of 1209.

Google's Gemini Omni 1.2 Flash Cuts AI Video Prices 30%

Google made Gemini Omni 1.2 Flash generally available on November 5. The video generation model now creates clips up to 15 seconds long (up from 10), supports conversational multi-turn editing, and costs $0.07 per second of 720p video — a 30% reduction from Omni 1.1.

The model accepts text, images, and short video as input, and outputs video with synchronized audio. It supports resolutions up to 4K. Every generated clip carries Google's SynthID watermark. The API model ID is gemini-omni-1.2-flash, available through Gemini API and Google Cloud.

For teams working with multi-modal prompting, the pricing drop and extended clip length make AI-generated video more practical for product demos, social content, and internal communications.

Thinking Machines Releases Inkling 2 Small

Mira Murati's Thinking Machines Lab introduced Inkling 2 Small on November 5 — a 594-billion-parameter Mixture-of-Experts model with 17 billion active parameters per token. It scores 31 on the Artificial Analysis Intelligence Index, more than double the original Inkling's 13. The model accepts text and image input and has a 1-million-token context window.

Inkling 2 Small is available now for inference and fine-tuning on Thinking Machines' Tinker platform. The company plans to release full weights under the Apache 2.0 license next week, pending external safety testing. The lab is also reportedly closing a $5 billion funding round led by Accel at a $40 billion pre-money valuation.

Open-Weight Models Hit by Private Test Shift

Related: Artificial Analysis updated its Intelligence Index v5.0 in late October to rely more heavily on private, previously unseen test sets. The result was a sharp widening of the gap between proprietary and open-weight models. Kimi K3.1 Pro, the top open-weight model, dropped from 55 to 46 — widening the gap from 3 points to 12 behind Gemini 4 Carbon's 59. This matters for teams making model selection decisions: reported benchmark parity between open and closed models may have been partly an artifact of test-set contamination.

AI Policy: DOJ Launches Super Intelligence Task Force

The U.S. Department of Justice announced the formation of its Super Intelligence Task Force (DOJ-SITF) on November 5, 2026. The task force is co-led by Associate Attorney General Stanley Woodward and Associate Deputy Attorney General Stephen Weiss, and it absorbs the DOJ's existing AI Task Force from January 2026.

The DOJ-SITF will coordinate criminal and civil enforcement against AI misuse by foreign state actors, transnational criminal organizations, and corporate actors. It will also recommend enforcement priorities, develop prosecution guidance, and serve as the DOJ's point of contact with the White House's Super Intelligence Force, chaired by Director of National Intelligence Jay Clayton.

Notably, the task force uses the Trump administration's "Super Intelligence" terminology per Executive Order 14434. For practitioners, the practical takeaway is the emphasis on corporate internal controls — the DOJ-SITF explicitly aims to ensure companies exercise "vigorous internal controls so that SI technology behaves as intended." Teams working on AI governance frameworks should note this enforcement signal.

DeepSeek Open-Sources a Proactive Coding Agent

DeepSeek's Harness team open-sourced dsh-superagent on October 31, an MIT-licensed research prototype of a proactive coding agent. Unlike typical coding agents that respond to explicit requests, this one monitors a GitHub repository and optionally team chat channels (GitHub and Feishu), watching for coding problems and offering solutions without being summoned.

The architecture uses a Router-Worker pattern: a long-running Router session handles incoming conversations and dispatches tasks, while isolated Worker sessions execute each task in its own directory. Proactive work defaults to "observe" mode — the agent logs what it would have done without acting, letting teams review before switching to auto mode.

This is a research prototype, not a finished product, but the pattern matters. It demonstrates how tool use patterns are evolving from request-response toward ambient, always-listening agent architectures.

Anthropic: Personal Finance, Project Tags, and IPO Updates

Anthropic shipped two Claude product updates this week. On November 5, Claude Personal Finance launched in beta for US Pro, Max, and Ultra users — you can link bank, credit card, loan, and investment accounts for read-only financial Q&A about your own spending, balances, and upcoming bills. On November 6, project tags arrived in beta for Enterprise organizations, letting teams tag projects with existing cost codes or client-matter numbers to track AI spend by project.

Meanwhile, Anthropic's amended S-1/A filing (November 2) added a disclosure that internal estimates suggest the company "may imminently meet" its Responsible Scaling Policy's threshold for AI-accelerated research — defined as AI that dramatically speeds up AI development itself. The filing notes that since April 2026, internal Claude usage at list prices exceeds payroll costs. Anthropic plans to list on Nasdaq under the ticker ANT, with the roadshow starting the week of November 10.

What This Means for Your AI Skills

This week's developments cluster around a clear theme: AI tools are shifting from answering questions to owning ongoing work. Perplexity Teammates, DeepSeek's proactive agent, and Cursor's Build Live Apps all push the boundary of what happens between human checkpoints. Understanding how to design, oversee, and evaluate these agentic workflows is becoming a core professional skill — exactly what the AI skills rubric maps.

The DOJ-SITF announcement reinforces that governance and compliance aren't optional extras. If you're deploying AI at work, your internal controls matter to regulators now, not just to your IT team.

Aisa's conversational AI reads this snapshot every week to stay current on the AI landscape, so when you take the assessment, your conversation reflects the latest developments — not a frozen training set.

Take the free AI skills assessment to see where you stand.

---

Related reading: Top 10 AI Skills Certifications for 2026 — Compare certifications and find the one that fits your career path.

Related reading: How Good Are Most People at AI? — See where your skills rank against thousands of other professionals.

Related reading: AI Fluency: The New Digital Literacy — Why baseline AI skills now matter for every role, not just tech.

Frequently Asked Questions

What are the biggest AI developments this week?

The biggest developments in the week of November 1-8, 2026 are Perplexity launching Teammates (persistent AI agents for $200/month), Cursor shipping Build Live Apps (prompt-to-hosted-web-app), and the U.S. DOJ forming its Super Intelligence Task Force for AI enforcement. OpenAI also released Images 2.6 with GIF support, and Google cut AI video pricing 30% with Gemini Omni 1.2 Flash.

Which new AI models launched this week?

Thinking Machines Lab released Inkling 2 Small, a 594-billion-parameter open-weight model scoring 31 on the Artificial Analysis Intelligence Index, with Apache 2.0 weights expected next week. OpenAI launched GPT-Image-2.6 Flare and Sunburst for image generation. Google made Gemini Omni 1.2 Flash generally available for video generation at $0.07 per second.

How do this week's AI changes affect professionals?

The shift toward persistent, always-on AI agents (Perplexity Teammates, DeepSeek's proactive coding agent) means professionals need to learn how to assign, monitor, and evaluate ongoing AI work — not just write single prompts. The DOJ's new task force signals that AI governance and internal controls are becoming enforcement priorities, making compliance skills increasingly important for anyone deploying AI at work. Take the AISA assessment to benchmark your readiness.