Perplexity AI
perplexity-ai-f914a629·7 events·first seen 28d agoAliases: Perplexity AI, Perplexity, perplexity
Co-occurring entities
More like this (12)
Recent events (7)
Perplexity production data shows AI agents perform 26 min of autonomous work vs 33 sec for search, cut task time 87%
A paper using production data from Perplexity's Search and Computer products quantifies how autonomous AI agents reshape knowledge work relative to conversational search. Key findings: Computer executes 26 minutes of autonomous work per session versus 33 seconds for Search, reduces task completion time from 269 to 36 minutes on matched tasks (87% time reduction, 94% cost reduction), and lowers per-query dissatisfaction by 55%. The study also finds agents shift user behavior toward higher-order tasks, cross occupational boundaries more often, and unlock work categories essentially absent from search usage.
Data Points: Perplexity Computer expands, Google Aletheia math agent, DeepSeek chip strategy, Nvidia retrieval pipeline, Stargate cancellation
The Batch's weekly data points roundup covers five significant AI developments: Perplexity expanded its Computer agentic platform to desktop, mobile, and enterprise with new APIs and financial data tools; Google released Aletheia, a Gemini-based math research agent achieving 95.1% on IMO-Proof Bench Advanced (up from 65.7%); DeepSeek withheld pre-release access to its V4 model from Nvidia and AMD while giving domestic Chinese chipmakers early access; Nvidia's NeMo Retriever topped the ViDoRe v3 leaderboard using a ReACT-based agentic retrieval loop; and OpenAI and Oracle cancelled plans to expand the Abilene Stargate campus from 1.2 GW to 2.0 GW due to financing and reliability issues.
SearchGPT: OpenAI Prototype for AI-Powered Search
OpenAI announced SearchGPT, a temporary prototype integrating real-time web search capabilities into a conversational AI interface. The prototype aims to deliver fast, timely answers with clearly attributed sources. It represents OpenAI's direct entry into AI-native search, competing with existing players like Perplexity and Microsoft Bing AI.
SubFit: Submodule-Level Fitted Residual Replacement for LLM Compression
SubFit introduces a post-training LLM compression method that operates at the submodule level (Attention and FeedForward separately) rather than full layers, and selects components non-contiguously. The approach replaces removed submodules with lightweight fitted residual bypasses calibrated on small data. Evaluated across ten LLMs at sparsity levels from 12.5% to 37.5%, SubFit retains 84.6% of dense downstream accuracy at 25% sparsity versus 81.6% for the strongest baseline, while reducing perplexity degradation from 4.34x to 2.42x and delivering measurable inference speedup and KV-cache savings.
State of Mobile 2026 Report: AI App Revenue Triples to $5B, Downloads Double to 3.8B
Sensor Tower's State of Mobile 2026 report documents explosive growth in mobile AI apps during 2025: global revenue tripled to over $5 billion and downloads doubled to 3.8 billion, with users spending 48 billion hours in AI apps — roughly 10x the 2023 figure. ChatGPT leads downloads, followed by Gemini, DeepSeek, Doubao, and Perplexity; OpenAI and DeepSeek together account for nearly 50% of global AI app downloads. Non-game app revenue exceeded gaming revenue for the first time, driven largely by AI spending. The data provides concrete evidence that AI assistant usage is becoming habitual and mainstream on mobile platforms.
Mistral AI Launches Major le Chat Update with Web Search, Canvas, Pixtral Large, and Image Generation
Mistral AI has announced a significant expansion of its le Chat assistant with several new capabilities in beta: web search with citations, a Canvas interface for collaborative document and code creation, multimodal document and image understanding powered by the new Pixtral Large model, and image generation via a partnership with Black Forest Labs (Flux Pro). The update also introduces shareable task agents for workflow automation and speculative editing for faster responses. All new features are currently offered on a free tier, positioning le Chat as a direct competitor to ChatGPT, Claude, and Perplexity.
Anthropic and AWS expand partnership with $4B investment and Trainium hardware collaboration
Anthropic announced an expanded partnership with Amazon Web Services, including a new $4 billion investment that brings Amazon's total stake to $8 billion, while establishing AWS as Anthropic's primary cloud and training partner. The collaboration includes deep hardware-software co-development on AWS Trainium accelerators, with Anthropic engineers writing low-level kernels and contributing to the AWS Neuron software stack to optimize model training from the silicon up. Claude on Amazon Bedrock is described as core infrastructure for tens of thousands of enterprises, with named deployments at Pfizer, Intuit, Perplexity, and the European Parliament. The deal also extends Claude's availability to AWS GovCloud and classified cloud regions for government customers.