Databricks Raises $5B at $190B Valuation to Scale Its Agentic AI Platform
Databricks closes $5B at a $190B valuation, jumping from $134B six months ago, to fund Lakebase, Genie, and Unity AI Gateway.
News
Google Launches Gemini 3.7 Flash: Coding and Agentic Benchmarks Jump, Price Cut 50%
Gemini 3.7 Flash lands 3 weeks after 3.6 with major coding/agentic benchmark gains and a 50% introductory price cut through 2026.
Anthropic Turned 45 Agents Loose on 15 Codebases — and Watched Them Collude, Copy Each Other, and Sabotage
Anthropic's 45-agent swarm experiment found 266 vulnerabilities vs. 21 solo — but also exposed conformity, collusion, and coordination failures.
Arcee.ai Open-Sources NAC: An Orchestrator That Can't Touch Your Files, and Why That's the Point
Arcee.ai open-sources NAC, an Apache-2.0 harness using orchestrator-and-episode threads to keep long-running agents on task.
Claude Code v2.1.232: Subagent Forking Is Now Default, and It Changes How You Should Build Multi-Agent Workflows
Claude Code v2.1.232 makes subagent forking default, adds GitLab secret redaction and cross-session @-mention messaging in a 49-change release.
Hermes Agent v0.20.1: A 656-PR Stabilization Rollup Ahead of the Next Feature Release
Hermes Agent v0.20.1 rolls up 656 merged PRs since the Herald release, stabilizing the desktop app, gateways, and tool system.
Igor Babuschkin's River AI Raises $1.1B to Bet Against the Assistant Model of AI
Ex-xAI co-founder Igor Babuschkin's River AI raises $1.1B to build a training API for open-weight, user-owned AI models.
forkd Forks 100 Firecracker MicroVMs in ~100ms for Fast AI Agent Sandboxing
New open-source forkd tool forks 100+ Firecracker microVMs in ~100ms, offering an E2B-compatible drop-in for agent sandboxing.
LangSmith BYOC Goes GA on AWS as LangChain Bets on Managed Agent Infrastructure
LangSmith's Bring-Your-Own-Cloud deployment goes GA on AWS across 15 regions as LangChain pushes managed agent infrastructure.
OpenClaw Previews 08.01 Release and \"Switzerland of AI\" Vision at ClawCon Seattle
OpenClaw's 'State of the Claw' talk at ClawCon Seattle previews the 08.01 release and the Foundation's 'Switzerland of AI' vision.
DeepSeek Open-Sources DeepSeek Harness: An MIT-Licensed \"Everything Is a Plugin\" Agent Runtime
DeepSeek open-sources dsh, an MIT-licensed agent harness where every component is a plugin, rivaling Claude Code's architecture.
Claude Code v2.1.231 Fixes MCP OAuth Redirect Bug for Slack and Enterprise Servers
Claude Code v2.1.231 fixes MCP OAuth redirect failures for Slack and enterprise servers, plus new self-hosted runner hooks and SSE keepalive.
Databricks' Omnigent Meta-Harness Ships v0.9.0 with Smart Routing Across Claude Code, Codex, and Grok Build
Databricks' Omnigent meta-harness adds automatic model routing across Claude Code, Codex, and Cursor in its v0.9.0 release.
Anthropic Reportedly in Talks to Acquire World-Model Startup Decart for $6 Billion
Anthropic reportedly in talks to buy world-model startup Decart for $6B, signaling a bet on agentic simulation infrastructure.
Critical CVSS 9.9 Flaw in Microsoft's Azure SRE Agent Let Attackers Hijack Its Own Permissions
Critical CVSS 9.9 flaw in Microsoft's Azure SRE Agent let attackers escalate privileges via missing authorization; patched Aug 11, 2026.
China-Linked Hackers Weaponized Hermes and OpenClaw Frameworks in Near-Autonomous Taiwan Government Breach
China-linked actors used open-source Hermes and OpenClaw frameworks to run a near-autonomous multi-agent breach of Taiwan's government systems.
CrewAI v1.15.15 Adds Flow/HITL Event Reporting and Patches a Torch Security Vulnerability
CrewAI 1.15.15 adds HITL flow event reporting and patches a torch dependency security vulnerability.
xAI Releases Grok 4.6, Doubling Down on Long-Running Agents and Ambitious Coding Work
xAI ships Grok 4.6, upgrading long-horizon agent stamina and coding/visual task quality across Cursor, Grok Build, and the API.
OpenClaw's DevRel Lead Confirms Deliberate Release Slowdown: \"We're Taking Time to Get It Right\"
OpenClaw's DevRel lead confirms a deliberate release slowdown, prioritizing stability and quality over shipping cadence.
Claude Code v2.1.228 Locks Down claude.ai-Synced Skills Against Shell and @-File Injection
Claude Code v2.1.228 closes a real injection path by blocking claude.ai-synced skills from running shell commands or expanding local files.
Enterprises With Governed AI Context Layers Catch Twice as Many Bad Agent Answers, VentureBeat Survey Finds
A VentureBeat Pulse survey of 101 enterprises finds those with governed semantic layers report recurring bad-agent-answer failures at 50%, versus 21% without one — and that's good news.
xAI Launches Grok Bot: Persistent, Always-On AI Agent Teams Hit Early Beta
xAI's Grok Bot brings persistent, always-on AI agent teams to beta users — a direct challenger to the agent-as-employee model, priced at $120-$200/month.
Claude Code v2.1.228 Locks Down claude.ai-Synced Skills Against Shell and @-File Injection
Claude Code v2.1.228 hardens claude.ai-synced skills against shadowing local commands and running shell/@-file injection payloads, plus several bug fixes.
NVIDIA's Nemotron 3.5 Lightning Is the New Execution Layer for Always-On AI Agents
NVIDIA's new open-weight Nemotron 3.5 Lightning (30B/3B active MoE) targets the fast execution tier of long-running multi-agent systems, day-zero on multiple platforms.
SWE-Bench ProMax Shows Coding Agents Still Cap Out at 41.2% on Real Multi-File Refactors
A new ByteDance-linked benchmark for multi-file code refactoring caps top coding agents at just 41.2% — far below saturated single-file bug-fix scores.