Perplexity (@Perplexity_AI) open-sources Lily, the local inference engine built for hybrid compute in Perplexity Computer. Lily is specialized for Qwen3.6-35B-A3B on Apple silicon so on-device compute does not bottleneck Computer tasks.
Together AI (@togethercompute) announces Equinix Inference Exchange with Equinix and NVIDIA: its open-model platform is live across Equinix’s global data centers, productizing both which model enterprises run and where it runs — arguing model choice and performance are not a trade-off.
Perplexity said Hybrid Compute is live for Pro, Max, and Enterprise subscribers in the Mac app on macOS 15+: cloud models handle search, planning, and reasoning while a local model processes private files and sensitive data, then merges into one answer without uploading that sensitive content. It gives builders and knowledge workers a hybrid agent path that can keep code and docs on-device.
GitHub said OpenClaw 2.0 is out and published a roundtable with creator Peter Steinberger and maintainers on how they built, scaled, and secured what it calls the fastest-growing open source project in GitHub history. OpenClaw is a device-local personal assistant that fans out to messaging channels and models through a gateway, so agent work lives next to existing chats rather than a new IDE. The 2.0 line is a large install and browser-app rewrite; for teams self-hosting a messaging-side agent, it is a checkpoint on both capability and the supply-chain load that viral agent repos now create.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.