2026
Gemini 3.1 Flash Live: Low-latency voice AI with native audio output
Asana's official Claude MCP connector can't create tasks in Claude Code CLI because its tools require interactive UI
Cohere Transcribe beats Whisper Large v3 on the HuggingFace ASR leaderboard with a 2B open-source model
Tencent's Covo-Audio is a 7B end-to-end audio model with full-duplex conversation via THINK, SHIFT, and BREAK tokens
Google's TurboQuant compresses LLM KV caches 6x with zero accuracy loss and near-zero indexing time
TinyLoRA fine-tunes a 7B model to 91.8% GSM8K accuracy with only 13 parameters using RL
GitAgent defines AI agents as git repos, exportable to any framework with one CLI command
NVIDIA's Nemotron-Cascade 2 hits gold-medal math and coding with a 30B MoE using only 3B active parameters
GLM-OCR: A 0.9B document parsing specialist beats 235B models
Baidu's Qianfan-OCR collapses the multi-stage OCR pipeline into one 4B model with Layout-as-Thought
Google's Colab MCP server lets any AI agent create and run notebooks in the cloud
NVIDIA's OpenShell enforces AI agent guardrails outside the agent process so a compromised agent can't override them
Unsloth Studio is an open-source no-code UI for training and running local LLMs
Mistral Small 4 merges instruct, reasoning, and coding into one model with per-request reasoning effort
OpenViking: A context database using filesystem paradigm for AI agents
You can force an LLM to only output valid answers
Pipe Mastra agent responses through jq to colorize reasoning and tool calls in the terminal
OpenClaw custom skills silently disappear without quoted YAML descriptions and openclaw metadata
A website can serve a full ANSI terminal UI when curled instead of HTML
Google DeepMind's Lyria 3 generates full songs from a photo or a sentence
HTTPie needs --ignore-stdin in non-TTY environments like Claude Code
HTTPie interprets numeric keys as array indexes, breaks Asana custom field updates
Claude Code runs natively as a GitHub Action
Kimi's agent swarm runs up to 100 parallel sub-agents to escape context window limits
glow strips color when called from a subprocess: force it with env vars
Zvec is an embedded vector database aiming to be the SQLite of vector search