NVIDIA Hit 100% on ARC-AGI-3 Without Building a New Model

In partnership with

No follow-up questions required

Every sales leader knows the feeling. You walk into a pipeline review with a number you believe in, and twenty minutes later, you're defending every line item to a CEO who just wants to know what's actually going to close.

HubSpot Sales Hub ends that conversation. Every deal, every rep's activity, and every buyer signal are all in one place and updated automatically. So your forecast is built on what's actually happening. And when you present that number, you can stand behind it.

Get a demo

NVIDIA Shows the Agent Harness Can Matter More Than the Model

NVIDIA published results for Agentic Variation Operators (AVO), a long-horizon agent architecture combining persistent memory, tools and a supervising agent. Using Claude Opus 5, AVO scored 100% on the public ARC-AGI-3 set, completing all 183 levels across 25 environments, compared with a roughly 30% model-level baseline. In a separate seven-day autonomous engineering run, AVO explored 500+ optimization directions, committed 40 GPU kernel versions and produced kernels up to 10.5% faster than FlashAttention-4.

The strategic shift: frontier-agent performance is becoming a systems-engineering problem, not simply a model-selection problem. Memory, supervision, tools and feedback loops may become as important as the underlying LLM.

Anthropic Puts Claude Mythos 5 to Work Defending Enterprise Code

Anthropic made Claude Mythos 5 available in Claude Security for Enterprise customers, where it can autonomously scan codebases for vulnerabilities and propose patches for human approval. Anthropic is also integrating Mythos into third-party cybersecurity products, expanding its Cyber Verification Program and launching a $35 million Defender Advantage Fund in Claude credits to secure open-source software. Importantly, users receive defensive artifacts such as patches without gaining unrestricted access to the powerful underlying model.

The strategic shift: Anthropic is demonstrating a potentially important deployment model for high-risk agents: expose narrowly scoped agent actions and outputs while keeping the most dangerous underlying capabilities behind controlled interfaces.

Slack Turns Coding Agents Into Multiplayer Teammates

Salesforce introduced Slack Code, dedicated code channels where teams can invoke coding agents, watch them work, review diffs, inspect live previews, redirect the agent and approve changes without leaving Slack. Founding integrations include Claude, Devin, GitHub Copilot, ChatGPT and Vercel agents, with Slack saying its marketplace already contains 500+ AI apps and agents. Code channels automatically create around a task and archive when the job is finished.

The strategic shift: coding agents are moving from private IDE sessions toward shared organizational infrastructure. Slack is positioning itself as the collaboration and governance layer where humans and multiple external agents work together.

The U.S. Army Wants Cyber Agents That Respond at Machine Speed

The U.S. Army detailed Project Griffin's Intelligent Response and Orchestration Node (IRON), an effort to use AI agents to ingest network-security signals and automatically execute defensive actions. Planned capabilities range from temporary firewall blocks to vulnerability patching, with adjustable confidence thresholds, audit trails, an emergency kill switch and the ability to undo autonomous actions. Army officials are also explicitly demanding lower token costs and safeguards against agents themselves becoming new attack surfaces.

The strategic shift: military buyers are moving beyond asking whether agents work. Procurement requirements are beginning to focus on the economics, auditability, security and controllable autonomy required to put agents into operational systems.

Solana Gets a Reconciliation Layer for Agentic Payments

ReconArt integrated its Recon Data Factory with Solana so enterprises can reconcile stablecoin transactions and payments initiated autonomously by AI agents alongside conventional financial data. The system targets the high-volume, real-time, machine-to-machine transactions that traditional batch reconciliation systems were not designed to process, while preserving audit, compliance and counterparty controls.

The strategic shift: the agentic payments stack is expanding beyond wallets and transaction execution. If agents begin making millions of autonomous microtransactions, enterprises will also need machine-native infrastructure for reconciliation, accounting, policy enforcement and auditability.

50 Excel Hacks That Make Spreadsheets Work For You

Most Excel users waste hours on tasks that take experts minutes. The difference is not talent. It is knowing the right shortcuts.

These 50 Excel hacks from Kenji Explains cover the functions, formulas, and workflows that make spreadsheets work for you instead of against you.

Each one comes with step-by-step instructions and ready-to-use templates. Subscribe to Marketing Against the Grain and get all 50 free.

Here's what you'll get:

• Time-saving shortcuts that eliminate the most common formula frustrations

• Interface essentials most users never discover on their own

• Game-changing functions for data analysis, automation, and visualization

• Guided practice template so you can apply every hack immediately

The people who are fast in Excel are not smarter. They just stopped fighting the tool.

Get 50 Excel Hacks free when you subscribe to Marketing Against the Grain today.

Get The Guide

🗺️ The Map of AI Agents (Live & Growing)

We’re charting the entire AI agent ecosystem — thousands of options across categories.
Your next agent is already on the map.

👉 Explore the Landscape Map →