Perplexity Open-Sources Lily, Its On-Device Inference Engine for Hybrid Compute
Perplexity has open-sourced Lily, the local inference engine powering hybrid compute in Perplexity Computer, optimized for Qwen3.6-35B-A3B on Apple silicon to k…
Global AI Digest
-
-
Muse Spark 1.3 Now Available on OpenRouter for Agentic Workflows
Muse Spark 1.3 is now available on OpenRouter, designed for long-running agentic, multi-agent, and coding workflows with learning, conflict handling, and clarif…
-
OpenRouter Launches Muse Spark 1.3 Contributor as Cost-Efficient Tier
Muse Spark 1.3 Contributor is now live on OpenRouter as a cost-efficient tier, alongside the updated Muse Spark 1.3 aimed at long-running agentic, multi-agent,…
-
Cline Completes Move to New SDK Harness, Reports Major Drop in Task Errors
Cline has migrated all Cline products to a new SDK harness built for open-weights models, with task mistake rates dropping from 6.34% to 0.62%.
-
OpenRouter Launches Muse Spark 1.3 Contributor: Cost-Efficient Tier for Agentic Workflows
OpenRouter has announced the Muse Spark 1.3 Contributor tier as the cost-efficient option, with the model now available for long-running agentic, multi-agent, a…
-
Muse Spark 1.3 Now on OpenRouter, Built for Agentic & Coding Workflows
OpenRouter has released Muse Spark 1.3, an updated model designed for long-running agentic, multi-agent, and coding workflows. It can track learned context, han…
-
MiniMax H3 Max Turbo Debuts on fal: 2x Speed, Half Cost, ~97% Quality
MiniMax's H3 Max Turbo, now available through partner platform fal, offers 2x faster video generation at half the cost while retaining about 97% of the original…
-
Command Code Hits $15M Run Rate, Claims Fastest-Growing AI Coding Agent Status
Command Code announced it has crossed a $15 million annualized run rate, citing 34,000 paying customers, 28 trillion tokens of scale, and a $10/mo AI coding pla…
-
Try Talking to an LLM to Build Confidence in Code Without Reading It
A practical exercise from a developer: instead of reading code directly, talk to an LLM and aim to reach confidence about what the code does. A useful tip for A…
-
Cline Migrates 11 Million Users to New Open-Weights SDK Harness, Cutting Mistake Rates to 0.62%
Cline moved all 11 million extension users to a new SDK harness built for open weights models, reducing task mistake rates from 6.34% to 0.62% in a major under-…
