ChatGPT Agent arrives – here's what it can actually do

Welcome to Lore Brief, your weekly edge in the age of AI.

This issue is brought to you by Factory, an engineer in every tab.

Factory, an engineer in every tab.

Factory is a Sequoia-backed startup that has created what I think is the best agentic coding platform out there. It’s better than Devin in all the tests I’ve done.

Whether you're prototyping quickly or building enterprise-grade solutions, Factory accelerates development. Major companies like Zapier, Bayer, MongoDB, and Framer use Factory to develop faster and solve more bugs, giving engineers time to focus on what they enjoy doing, not debugging.

67 K-view demo: watch me and founder Matan Grinberg clone the core DocuSign functionality in 15 minutes!

Try Factory at Factory.ai and follow Factory on Twitter/X for more updates!

OpenAI Unveils ChatGPT Agent: A Leap Toward Autonomous AI

OpenAI launched ChatGPT Agent yesterday, combining web browsing, computer control, and code execution into one new product. The new "Agent mode" merges capabilities from Deep Research and Operator into a single tool that operates through its own virtual computer.

Core capabilities:

• Text browser - Reads web content and conducts research

• Visual browser - Sees and clicks on websites like a human user

• Terminal - Runs code, manipulates files, creates documents

The agent handles complex multi-step workflows: research competitors and build PowerPoint presentations, analyze data and generate spreadsheets, log into websites to gather information, work autonomously for 10+ minutes while users focus elsewhere.

Agent mode is available today but with usage limits: Pro users get 400 messages monthly at $200, Plus users get 40 at $20.

AI "agents" have been Silicon Valley's favorite buzzword for the last year. ChatGPT Agent is OpenAI's first serious attempt to move beyond smart responses to autonomous task completion - the digital employee shift everyone's been predicting.

The promise: AI becomes your delegated employee, not just a chat and a search engine. You assign tasks and projects, get deliverables.

The reality: This is early-stage tech requiring significant oversight. OpenAI warns about prompt injection risks and potential security vulnerabilities. The limited usage allowances signal this will evolve gradually over 6-12 months, not transform work overnight. Anyone doing knowledge work should pay attention, but expect gradual adoption rather than immediate revolution.

The Next Wave: How Tom Bilyeu Uses AI + Why He'll Never Hire Again

Thinking Machines Lab Secures $2B Funding as Launch Approaches

Mira Murati's startup Thinking Machines Lab raised $2 billion in a funding round led by Andreessen Horowitz, achieving a $12 billion valuation just months after its February founding. The investment supports development of collaborative general intelligence through multimodal AI systems, with initial products expected soon.

• Founded by former OpenAI CTO Mira Murati, the company focuses on safer, scalable AI with an emphasis on open-source elements.

• Investors include NVIDIA, Accel, ServiceNow, Cisco, and AMD, highlighting strategic partnerships in hardware and enterprise tech.

• No revenue or products yet, but the round reflects high confidence in Murati's vision for multimodal models that integrate text, image, and other data.

• Plans involve building AI that empowers humanity via collaborative intelligence, potentially including tools for creative and scientific applications.

• The massive seed round sets a record, surpassing previous benchmarks in AI startup funding

This influx of capital into a nascent venture signals intense investor faith in proven leaders like Murati, who bring expertise from top labs to tackle AGI challenges. It also reveals how hardware giants are betting on software innovations to drive demand for their chips.

Early-stage bets like this could spark a wave of specialized AI firms, diversifying the field beyond a few dominant players.

Zuckerberg Outlines Meta's Vast AI Supercluster Initiative

Mark Zuckerberg announced Meta's strategy to construct enormous AI data centers, committing many billions in spending. The plan features superclusters like Prometheus, slated for 2026 activation, to fuel pursuits in superintelligence.

• Investments target multi-gigawatt facilities, with Prometheus at 1 GW and Hyperion at 5 GW, rivaling city-scale power usage.

• Focus lies on infrastructure for training advanced models, including personal superintelligence for users.

• Energy demands prompt exploration of nuclear and renewable sources to sustain operations.

• This builds on Meta's 2025 capex of $60-65 billion, up from prior years, amid competition with OpenAI and Google.

Meta's aggressive infrastructure push reflects the resource intensity of frontier AI, where compute power dictates leadership. It highlights Zuckerberg's shift toward hardware dominance to support open-source efforts like Llama.

Scaled facilities will accelerate model training, enabling more sophisticated assistants integrated into social platforms. This infrastructure race may lead to energy innovations, but it raises questions about sustainability in an AI-driven economy.

Cognition AI Absorbs Windsurf Following Google's Talent Raid and Licensing Pact

Cognition AI, developers of the Devin coding agent, completed its purchase of competitor Windsurf, incorporating the firm's intellectual property, workforce, and user community. The acquisition occurred shortly after Google secured a $2.4 billion arrangement to onboard Windsurf's founders and access its technology.

• Windsurf's agentic IDE complements Cognition's autonomous agents, promising enhanced developer tools.

• The move followed OpenAI's expired $3 billion offer, illustrating heated competition for AI coding startups.

• Integration aims at breakthrough experiences, combining Windsurf's platform with Devin's capabilities.

• Valued as a strong outcome for employees, it consolidates talent in a fragmented coding AI space.

Rapid consolidations like this expose the cutthroat nature of AI talent acquisition, where big tech cherry-picks expertise while smaller firms merge to survive. It points to coding agents as a hot sector, with Devin-like tools evolving through such deals.

Combined strengths could yield more reliable AI programmers, streamlining software creation for teams. As mergers proliferate, they might concentrate innovation, prompting calls for policies that nurture diverse startups.

Lovable AI Startup Achieves Unicorn Status with $200M Funding

Swedish startup Lovable raised $200 million in Series A on July 17, valuing it at $1.8 billion and making it Europe's newest AI unicorn. The company specializes in vibe-based coding tools powered by AI.

• Funding led by investors betting on intuitive programming interfaces.

• Lovable's platform uses AI to interpret user intent for code generation.

• Round highlights Europe's growing AI scene amid global competition.

• Startup plans expansion into enterprise tools and developer ecosystems.

• Valuation reflects optimism in AI-assisted software development.

This funding wave validates niche AI applications, showing investor appetite for tools that simplify complex tasks.

Moonshot AI's Kimi K2 Claims Top Spot in Open-Source Rankings

Moonshot AI rolled out Kimi K2, a trillion-parameter mixture-of-experts model excelling in benchmarks for coding, reasoning, and long-context tasks. Backed by Alibaba, this open-source release positions it as a leading alternative to proprietary systems like GPT-4.

• With 32 billion active parameters, it scores high on SWE-bench (65.8% accuracy) and other metrics, surpassing rivals in agentic behavior.

• Optimized for software development, it handles large-scale training stably at low costs.

• Available via GitHub, it supports fine-tuning and deployment for diverse applications.

• Focus areas include extended context windows and multimodal integration for real-world use.

• As China's contribution, it challenges US dominance in open AI models.

Kimi K2's performance metrics affirm China's strides in AI, offering high-capability open models that reduce dependency on closed systems. Its efficiency in parameters highlights advances in mixture-of-experts architecture for practical deployment.

Runway Unveils Act-Two for Advanced Video-to-Animation Conversion

Runway introduced Act-Two, an enhanced AI system that transforms footage of individual performances into detailed animated figures. This tool handles tracking for heads, faces, bodies, and hands while adapting to different artistic approaches and formats.

• The model improves generation quality over prior versions, adding environmental motion to static character inputs automatically.

• It works with various angles, including non-human subjects, making it versatile for filmmakers and animators.

• Users can input a single performance video alongside a character reference to produce full animations with background elements.

• Available through Runway's platform, it supports real-time creation and integration into creative workflows.

• Demonstrations show major leaps in realism, positioning it as a step toward democratizing high-end motion capture.

Releases like Act-Two demonstrate steady progress in generative video tech, where accessibility meets professional-grade output to lower barriers for creators.

Tools of this caliber will likely expand into gaming and virtual production, reducing costs for independent artists. Broader access could fuel a surge in personalized content, altering how stories get told across media.

That’s it for today.

Consider forwarding Lore Brief to a colleague to help them get ahead in the AI Age.

-Nathan Lands
Connect → X | LinkedIn
Listen to The Next Wave → Apple | Spotify | YouTube

(Disclosure: I may own equity in companies mentioned in Lore Brief.)