Perplexity Pushes Multi-Agent Orchestration With Computer

Perplexity Pushes Multi-Agent Orchestration With Computer

While OpenAI grabbed the DevDay headlines, Perplexity has quietly been building the same future from a different angle. Computer, its agent-of-agents system, now has Automations and Scheduled Tasks — recurring background workflows with triggers, app integrations, and sub-agents that report up to a coordinating layer [4][5]. Claude Opus 4.6 sits at the core, but the real story is orchestration: Computer doesn't just do the work, it assigns the work.

This is agent governance made practical rather than theoretical. Persistent monitoring, document creation, tool use — all running autonomously in the cloud for Pro/Max/Enterprise subscribers [6]. The framing matters: Perplexity is explicitly building toward multi-agent systems with safeguards as a selling point, not an afterthought bolted on after regulators complain.

The X conversation around this has centered on exactly that tension — how do you get the productivity gains of autonomous sub-agents without losing the ability to audit what they did and why. Nobody's fully answered it yet, but the fact that it's the central question says something about where the industry's head is at.

Pearson Buys Workera to Answer the Question Everyone's Avoiding: Can Your People Actually Use This?

Pearson announced on September 29 it's acquiring Workera, the Palo Alto-based AI-native skills assessment platform that uses agentic AI and adaptive simulations to verify what employees can actually do, not just what they claim on a resume [7]. Workera already counts ServiceNow, Accenture, and the US Space Force as clients, and its ~60-person team — including CEO Kian Katanforoosh — is moving into Pearson's leadership structure, with the deal expected to close in H2 2026 [8][9].

Andrew Ng, Workera's board chair and Coursera co-founder, framed this as mission alignment: as AI reshapes what skills matter, you need a rigorous way to measure them, not just a certificate mill [7]. That's the quiet but important story here — every company shipping agentic tools this week is also creating a skills gap, and almost nobody has a real answer for closing it.

This deal is a bet that "does your workforce know how to work with AI" becomes as standard a corporate question as payroll compliance. Given how fast Dots, Computer, and their successors are moving, that bet looks safe.

EU AI Act Delays Are Quietly Reshaping Who Gets Access First

The EU approved amendments in June 2026 pushing back high-risk AI obligations — standalone systems to December 2027, product-embedded systems to August 2028 [10]. Combined with GDPR, that regulatory stack is now visibly slowing feature rollouts: OpenAI's Dots is limited or delayed for Pro users in the EEA and UK, even as business and enterprise tiers see less friction in some markets [11]. Studies cited put over 11% of LLM releases delayed in the EU relative to the US specifically due to regulatory factors [12].

Call it Europe's double bind: the caution is defensible on privacy and safety grounds, but the practical effect is that European consumers and smaller businesses are increasingly the last to get access to frontier agentic tools, while enterprise customers with legal teams and compliance budgets get there faster. That's a widening gap, not a temporary one, and it's worth Nordic companies watching closely — the rules that protect you can also be the rules that leave you behind.