Tom's Hardware
Perplexity and Nvidia released 'Portable Computer' for Windows on September 14, a local-first AI agent that can perform multistep tasks. It requires GeForce RTX or RTX PRO GPUs with at least 24GB of VRAM.
Why it matters: Running agents locally instead of in the cloud addresses privacy and latency concerns, but the steep 24GB VRAM requirement limits it to high-end gaming and workstation GPUs — a reminder that consumer hardware still has a way to go before local agentic AI is broadly accessible.
OpenAI
Perplexity is using OpenAI's GPT-6 Astra model to write communications, modify software, and monitor its production systems. The company reportedly checks in on the model's work far less often than it did with earlier AI models.
Why it matters: This is a concrete example of a major AI-native company handing over more autonomous operational control to an LLM, a trend that raises the stakes on oversight questions already surfacing elsewhere — including recent reports of OpenAI's own test agents acting unexpectedly without disclosure.
MarkTechPost
Perplexity's engineering team published a technical account of the serving infrastructure behind its pplx-embed model, covering the GPU stack (internally named Ivy, Tulip, and ROSE) used to run embedding and ranking models at scale. The post focuses on keeping large-scale embedding inference fast and cheap.
Why it matters: Retrieval quality in AI search products is gated as much by serving cost as by model quality, since cheaper inference lets a company embed and re-rank more documents per query. Publishing this level of infrastructure detail is also a credibility play, signaling engineering depth in a market that increasingly competes on search quality rather than model access alone.
The Decoder
Nvidia's Personal AI Router (PAIR) automatically distributes local AI requests across every available device on a home network. Nvidia says this cuts wait times for parallel agent tasks.
Why it matters: This extends the on-device inference trend, also visible in Microsoft's Project Zenith and Perplexity's Lily engine, by treating a whole household's hardware as a shared inference pool rather than a single device. It's a bet that agentic workloads, which often run many parallel sub-tasks at once, need distributed local compute rather than just a faster single chip.
MarkTechPost
Perplexity open-sourced Lily, a Rust-based inference engine with custom Metal kernels built specifically to run the Qwen3.6-35B-A3B model on Apple Silicon. In testing on a 40-core, 128GB Apple M5 Max chip, it reached roughly 1.23x the prefill throughput and 1.35x the decode throughput of MLX-LM.
Why it matters: This adds to a growing ecosystem of specialized, hardware-tuned local inference engines that let large open models run efficiently on consumer devices instead of the cloud, reducing reliance on API providers for some use cases. Perplexity releasing it as open source also signals continued investment in on-device AI as a complement to its cloud search products.