WireTensors

Updated Sat, 27 Jun 2026 21:31:57 UTC

Perplexity's Always-On AI, Qualcomm's $12B Chip Bet, and the New Era of Agentic AI — 27 June 2026

Roundup facts
Published 2026-06-27
Items 6
Coverage Writing, coding, image, video, productivity, SEO
Last verified 2026-06-27

Perplexity's Personal Computer: AI That Never Sleeps

Perplexity has released a software system—not hardware—that runs continuously in the background on compact machines like a Mac Mini, operating as a persistent 24/7 AI agent. Rather than waiting for you to ask questions, it executes tasks autonomously, suggesting a fundamental shift in how people will interact with AI. This matters because it breaks the 'prompt-and-wait' model that has dominated consumer AI since ChatGPT's launch. Early adopters are already exploring what persistent agents can do when given long stretches of unattended time; this could reshape expectations around productivity tools and personal automation.

Qualcomm's $12 Billion AI Infrastructure Offensive

In what amounts to a strategic repositioning of the chip giant, Qualcomm announced $8–10 billion in acquisition talks with Tenstorrent (an AI chip company) and separately confirmed a $3.9 billion acquisition of Modular, an AI software firm specialising in optimising AI runtimes and compilers. Together, these moves signal Qualcomm's determination to compete head-to-head with NVIDIA in next-generation AI hardware and software integration. For enterprises and AI builders, this could mean faster, cheaper inference and new alternatives to NVIDIA's dominant position—but the consolidation also tightens the supply chain for cutting-edge AI infrastructure during a period of geopolitical tension over chip access.

Nvidia Nemotron 3 Super and the Multi-Agent Turn

Nvidia released Nemotron 3 Super, a 120-billion-parameter reasoning model explicitly built to power multi-agent AI systems. This reflects a industry-wide shift away from single-model inference toward architectures where multiple specialised agents collaborate on complex tasks. The model is designed for enterprise and infrastructure use, not consumer chat—meaning we're entering a phase where the sophistication of AI systems depends on how well different agents coordinate. This matters for builders; it's a signal that the competitive frontier has moved beyond raw model size toward orchestration and reasoning across distributed agent teams.

Gemini's Multimodal Embedding and Video Generation Go Live

Google shipped Gemini Embedding 2, which converts text, images, videos, audio, and PDFs into unified mathematical vectors for search and analysis, alongside Gemini Omni Flash, a model that generates video directly from any input grounded in real-world physics. These are now available to Google AI Ultra and Pro subscribers. The significance lies in breaking down the barriers between modalities—one model can now ingest and reason across documents, audio, and visual content simultaneously, making it practical to build applications that operate across what were previously separate domains (search, video creation, content analysis). This is the infrastructure Google needs to compete with OpenAI's multimodal and agentic capabilities.

ChatGPT Gets Real-World Agency (and New Limits)

OpenAI has enabled ChatGPT users to pause long-running queries, add context, and trigger real-world actions—browsing the web, sending emails, executing code—without restarting the conversation. Simultaneously, reports suggest GPT-5's full launch has been delayed at U.S. government request and restricted to vetted partners only, as has Anthropic's Claude Mythos 5. This creates a paradox: AI agents are becoming more capable and autonomous, yet regulatory gates are tightening around the most advanced models. The pragmatic impact is that developers will build agents against constrained, approved models for longer than expected, creating an unusual bottleneck between capability and deployment.

Snapchat Commits $400M/Year to Perplexity as Default AI

Snapchat has agreed to make Perplexity its standard AI assistant across the platform starting January 2027, under a $400 million annual deal. This is significant not because Perplexity is displacing OpenAI or Google, but because it signals a real-world deployment of third-party AI at massive scale. Snapchat's younger user base will now interact with Perplexity's search-centric reasoning model as their primary AI interface. For the AI industry, this validates Perplexity's positioning and shows that distribution partnerships—not just model capability—determine which AI systems reach users.

Roundup FAQ

What is this roundup? +

Perplexity launches a 24/7 AI agent for your Mac, Qualcomm moves aggressively on AI infrastructure with two major acquisitions totalling $12 billion, and the industry is quietly pivoting toward multi-agent reasoning systems that work without human prompts. The competitive landscape is shifting fast.

When was it published? +

This roundup was published and verified on 2026-06-27.

What topics does it cover? +

It covers: Perplexity's Personal Computer: AI That Never Sleeps; Qualcomm's $12 Billion AI Infrastructure Offensive; Nvidia Nemotron 3 Super and the Multi-Agent Turn; Gemini's Multimodal Embedding and Video Generation Go Live; ChatGPT Gets Real-World Agency (and New Limits); Snapchat Commits $400M/Year to Perplexity as Default AI.

Is the coverage neutral? +

Yes. Roundups summarise developments neutrally and do not promote any single vendor.

Does this roundup contain affiliate links? +

Links within roundups may be affiliate links; we may earn a commission at no extra cost to you, and this never affects coverage.

Where does the information come from? +

Roundups summarise vendor product pages, changelogs and public announcements, each verified on the publication date.

How often are roundups published? +

WireTensors aims to publish short roundups on a daily cadence.

Where can I read full tool reviews? +

Each tool mentioned has a full review under /tools, with pricing, ratings, pros, cons and FAQs.

Reviewed by Arjun Mehta

AI tools analyst; 8+ years reviewing SaaS and developer tooling

Last verified:

Sources