Inco Splash is an open-source inference engine built for Apple silicon. It claims up to 3× the decode speed of Ollama, 2× oMLX, and almost 4× when an agent fans out into sub-agents. Qwen3.8-27B runs at 144 tokens per second on an M5 Max MacBook Pro.
Key points
Qwen3.8-27B at 144 tok/s on an M5 Max MacBook Pro
Up to 3× the decode speed of Ollama, 2× oMLX, and almost 4× with sub-agents
Sources
- Original post
- discord.gg - Linked in the post
Community plugins – surpassed 1000 today in marketplace
Community plugins surpassed 1000 today, announced by the team. A link to view all plugins is provided.
Key points
Milestone: 1000 community made plugins
Link to view all plugins in the marketplace
Sources
- Original post
- hermes-agent.nousresearch.com - Linked in the post
Orca build workflow creates 9 PRs in morning
Orca build was used to orchestrate agents and generate nine pull requests ready for review by morning, after planning a Lifetime deal with Claude and reviewing with Codex. The workflow split work into incremental PRs and used Orca’s orchestration tool for agent communication.
Key points
Orca uses its orchestration tool for agents to talk to each other.
Nine PRs were ready for review in the morning, and a new product in Test mode was created in creem_io.
Sources
Task Launcher – side panel adds custom prompts per project
Task Launcher adds a side panel in Orca that lets you set a custom prompt for each project and launch a task with a button click. It runs GitHub issues as branches and worktrees from the panel. Installation is through Orca Settings → Plugins using the community catalog.
Key points
Custom prompt per project set in the side panel
Clickable button triggers a task run
Runs GitHub issues in separate branches and worktrees
Install via Settings → Plugins → Add marketplace URL then search for Task Launcher
Sources
- Original post
- Task Launcher - Linked page
Zone Log launched on App Store – free auto time tracker
Zone Log launched on the App Store as a free app that automatically tracks time spent at saved locations. It has no ads or subscriptions and the author seeks feedback.
Key points
Launch: Available on the App Store.
Pricing: 100% free, no ads, no subscriptions.
Sources
- Original post
- apps.apple.com - Linked in the post
DeepSeek V4.1 Flash and Nebius models now on Eden AI
DeepSeek V4.1 Flash and other Nebius models are now available on Eden AI. Reusing KV cache across turns cuts LLM inference costs at scale.
Key points
Reusing KV cache across turns cuts LLM inference costs at scale
DeepSeek V4.1 Flash and other Nebius models are on Eden AI
Sources
- Original post
- edenai.co - Linked in the post