Built on its Kilo Code acquisition, Kilo Desktop lets a lead agent split work across subagents that pick their own models from 500+ options (local ones included), with notebook sessions where you and the agents edit and run cells together. Anaconda also says its Enkrypt team found vulnerabilities in 73% of agent tools across 25,000+ MCP servers, so vet the servers you plug in.
As of Oct 6, paid API orgs sit in Build, Launch, or Grow, and you reach Grow, the new highest tier, after $500 in total API payments instead of the old $1,000 for the top tier. Existing paid orgs move automatically, so check your org's limits page to see your new rate limits before you plan around them…
The expanded Claude for Startups program, announced Oct 6 at SF Tech Week, includes up to five premium Claude Team seats for a year, $1,000 in API credits, Claude Marketplace access for building plugins, and office hours with Anthropic's Applied AI team. You qualify if your company was founded in the last five years or raised funding in the last two, and you apply through the Claude for Startups page…
Google DeepMind's Apache 2.0 EmbeddingGemma 2 (740M parameters, or 270M for text only) has an 8K context, 768 dim vectors you can truncate to 128, and jumps from 68.76 to 78.68 on MTEB Code, so it's a fit for local codebase indexing and coding agent retrieval. Weights are on Hugging Face and Kaggle and run in sentence transformers 6.1+, Ollama, llama.cpp, vLLM, MLX, and LM Studio…
Announced Oct 6 at Transcend, goal driven flows (GA this month) take an objective from the Duo CLI, headless mode, or Agentic Chat and carry it through review, tests, security scans, and approvals, while GitLab hosted GLM 5.3, Kimi K3, and MiniMax 3 are GA now with up to 8x more model calls per credit than frontier pairings. The MCP Server for outside coding agents also goes GA this month, and new credit caps can auto pause AI spend at a limit you set…
The Oct 5 release adds vllm preload , a daemon that keeps post quantized weights resident in GPU memory so engine restarts skip reloading, plus a big DeepSeek V4.1 Flash speed pass on Blackwell (SM100). Check your configs before upgrading: per request mm processor kwargs and media io kwargs are now rejected unless you pass trust request mm kwargs , tokenizer mode="slow" is gone, and online quantization="fp8" becomes fp8 per tensor…
Spotted something we missed? Start a thread.