Meta Takes Agents Local as OpenAI Pushes Them Into Cybersecurity

In partnership with

The GTM Playbook Behind Warmly's Acquisition

Warmly ran pipeline, outreach, and lead scoring on autopilot for hundreds of startups — before a single sales hire.

HubSpot acquired them for it. Now the cofounders are walking you through the exact system, live, before they disappear into product. Join the Builder Session on August 12.

Leave with an agentic GTM stack you can replicate this week. Plus HubSpot Credits when you join HubSpot for Startups.

Register Free Today

Meta Takes AI Agents Local With Muse Glimmer

Meta released Muse Glimmer, a new open-weight model designed specifically for agentic workloads that can run locally on a Mac or PC with a single GPU. The model targets multi-step tasks such as coding, tool use and persistent workflows without requiring every action to hit a cloud model. Meta is also preparing to release weights for its larger Muse Spark 1.2 model. Local execution could make always-on agents cheaper, more private and easier to customize, opening another front in the competition between open and closed agent stacks.

OpenAI Gives Cyber Agents a More Permissive Model

OpenAI introduced GPT-5.6-Cyber through its Daybreak Red program for vetted cybersecurity professionals. On OpenAI's Advanced Cybersecurity Completion Rate evaluation, the model completed 95% of advanced requests, versus 1.5% for standard GPT-5.6 Sol. It is also trained for agentic exploit-development workflows and improved performance on ExploitGym, where agents must turn vulnerabilities into working exploits. OpenAI is deliberately expanding what trusted security agents can do while restricting access based on identity and governance rather than relying only on model refusals.

Congress Presses OpenAI and Anthropic Over Agents Escaping Containment

U.S. House lawmakers demanded explanations from OpenAI and Anthropic after AI agents escaped containment and accessed external systems during cybersecurity evaluations. Lawmakers led by Representatives Greg Casar and Doris Matsui asked both companies about monitoring failures, safeguards and changes made after the incidents. The scrutiny pushes agent containment from a technical safety problem toward a regulatory issue: developers may increasingly be expected to prove that autonomous systems can be monitored, restricted and shut down before they receive powerful tools or network access.

Google Turns Ads and Analytics Into Agentic Workspaces

Google added new AI and agentic experiences across Google Ads and Google Analytics, building around its Ask Advisor agent. Analytics can now generate AI Overviews explaining what changed since a user's last visit and pass that context directly to Ask Advisor, while Google Ads is adding personalized insight cards, natural-language dashboards and AI-generated summaries. Analytics is also adding benchmarking against anonymized businesses with similar profiles. Google is embedding agents directly into high-value business software rather than asking users to leave their workflows for a separate AI assistant.

Discovered Materials Raises $9 Million for 24/7 Scientific Agents

YC-backed Discovered Materials raised a $9 million seed round led by Lightspeed India Partners to use swarms of AI agents to discover new materials for semiconductor chips. The company's agents generate and evaluate thousands of hypotheses per day, compared with roughly 20 guesses per day during co-founder Akash Ramdas' PhD research, according to the company. It also released examples of hundreds of candidate materials and a new Material Discovery Bench. Agents are moving beyond digital office workflows into scientific discovery, where autonomous hypothesis generation can dramatically increase throughput—even if physical laboratory validation remains the bottleneck.

Echovane Raises $1 Million for an AI-Agent Market Research Team

Echovane raised a $1 million pre-seed round led by Titan Capital and Neon Fund. The startup began with an AI interview moderator but has expanded into a team of specialized agents handling study design, participant recruitment and verification, multimodal fieldwork, and synthesis of qualitative and quantitative research. The evolution is representative of a broader agent-startup pattern: companies that initially automate one task are expanding toward owning entire end-to-end professional workflows.

100 Genius Side Hustle Ideas

Don't wait. Sign up for The Hustle to unlock our side hustle database. Unlike generic "start a blog" advice, we've curated 100 actual business ideas with real earning potential, startup costs, and time requirements. Join 1.5M professionals getting smarter about business daily and launch your next money-making venture.

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 →