Ideogram Vector 1.0: Generate Editable SVGs via API
Ideogram Vector 1.0 ships two API endpoints that generate real, editable SVGs from text prompts and images. Here is what the output contains, ...
Databricks MemCurrent: Agent Memory That Triples Pass Rates
Databricks MemCurrent ships in beta: a two-layer agent memory architecture that raises task pass rates ...
Cloudflare security-audit-skill: AI Agent Security Auditor
Cloudflare's open-source security-audit-skill is trending #1 on GitHub this week. Here's how its adversarial 6-phase ...
Flint 40B: Recursive’s Coding Model Was Built Mostly by AI
Recursive Superintelligence's Flint is a 40B MoE coding model that runs at 125 tok/sec locally ...
LiteLLM 1.107.0 Supply Chain Attack: Check Your Keys Now
LiteLLM 1.107.0 was trojaned on October 21, 2026. The credential stealer runs on every Python ...
VS Code 1.143: Copilot Default and Local AI With a Catch
VS Code 1.143 makes Copilot the default agent harness, ships experimental local AI inference with a 96GB GPU requirement, and adds multi-model comparison.
IQuest-Q1: The 320B Open-Weight Model for CLI Agents
IQuest Research released IQuest-Q1, a 320B sparse MoE with 512K context and Claude Code integration. ...
Kimi K3.1 Pro: The Open-Weight Frontier Narrows to 3 Points
Moonshot AI released Kimi K3.1 Pro, a 2.8T open-weight model that narrowed the frontier gap ...
NVIDIA OpenShell: Run AI Agents Without Handing Them Your Keys
NVIDIA OpenShell is a kernel-level sandbox runtime that runs AI agents in deny-by-default environments where ...
Agensh: Microsoft Ran 1,024 Agents with No Boss and Won
Microsoft Research open-sourced Agensh, a multi-agent harness that runs 1,024 coding agents without a central ...