- Luis Poveda's AI Newsletter
- Posts
- Luis Poveda's AI Newsletter: May 19, 2025
Luis Poveda's AI Newsletter: May 19, 2025
AI Agents Evolve While Infrastructure Expands
Luis Poveda
May 19, 2025
In partnership with
There's nothing artificial about this intelligence
Meet HoneyBook—the AI-powered platform here to make every client relationship more productive and prosperous.
With HoneyBook, you can attract leads, manage clients, book meetings, sign contracts, and get paid.
Plus, HoneyBook AI tool summarizes project details, generates email drafts, takes meeting notes, predicts high-value leads, and more.
Executive Summary
This week in AI showcases significant developments across the industry, with major players advancing autonomous agent capabilities while tackling inherent challenges. HuggingFace launched its Open Computer agent for free web-based automation, while OpenAI released Codex, bringing powerful coding capabilities to ChatGPT users. On the infrastructure front, CoreWeave inaugurated a cutting-edge Barcelona datacenter, and Microsoft announced substantial engineering layoffs amidst AI investment pressures. Creative AI saw advancement with Manus AI's innovative task-oriented image generation approach, while a Wall Street Journal report highlighted persistent obstacles for AI agents. These developments collectively signal a shift toward more autonomous, capable AI systems with increasing infrastructure investment and specialization.
AI Agents & Automation
HuggingFace Launches Open Computer Agent
HuggingFace has introduced a free, cloud-hosted AI agent called Open Computer Agent that can operate a virtual computer using text instructions. Unlike proprietary alternatives, this agent utilizes a Linux virtual machine preloaded with applications like Firefox and responds to natural language prompts, demonstrating the increasing capabilities of open AI models for complex workflows. While impressive, users should note that the agent can be somewhat sluggish and occasionally makes mistakes.
HuggingFace
"As vision models become more capable, they become able to power complex agentic workflows."
OpenAI Launches Codex Platform for AI-Assisted Coding
OpenAI announced the release of Codex, its most powerful AI coding agent to date, now available in ChatGPT. Powered by codex-1, a version of the company's o3 AI reasoning model optimized for software engineering, the agent runs in a sandboxed virtual computer and can handle complex coding tasks including writing features, fixing bugs, and running tests. Available initially to ChatGPT Pro, Enterprise, and Team subscribers, Codex is OpenAI's latest move in the competitive AI coding assistant market.
"We're about to undergo a pretty seismic shift in terms of how developers can be most accelerated by agents".
Browser Use Introduces Workflow Use Browser Automation
Browser Use has launched Workflow Use, an innovative browser automation solution that moves beyond traditional automation approaches. Workflow Use allows users to create deterministic workflows by recording their actions once, after which the system automatically generates executable workflows that can be run repeatedly. The tool features intelligent noise filtering from recordings, automatic variable extraction from forms, and an AI fallback system that activates when a workflow step fails, making it particularly resilient compared to traditional browser automation tools.
Workflow Use
“WorkflowUse is the easiest way to create and execute deterministic workflows with variables which fallback to Browser Use if a step fails."
AI Agents Face Significant Obstacles According to WSJ
A recent Wall Street Journal report highlights the substantial challenges still facing AI agents despite their rapid advancement. While multiple sources indicate 2025 is shaping up to be "the year of the AI agent," practical implementation still faces hurdles in reliability and consistency. According to IBM experts, while agentic systems will increasingly augment human work, the industry view is shifting toward a human-in-the-loop approach rather than complete autonomy, with complex tasks still requiring human oversight and decision-making.
WSJ
"You're still going to have cases where as soon as something gets more complex, you're going to need a human".
Infrastructure & Corporate Developments
CoreWeave Opens New Barcelona Datacenter
CoreWeave has officially inaugurated a state-of-the-art data center in Barcelona, Spain, in partnership with MERLIN Edged. The facility will host one of Spain's first large-scale NVIDIA Hopper training and inference supercomputers, equipped with over 10,000 of NVIDIA's most advanced AI chips. Strategically located at the Port of Barcelona, the data center offers 15 MW of critical load capacity powered by 100% renewable energy, reinforcing Barcelona's position as a key European tech hub and expanding CoreWeave's European AI infrastructure footprint.
MERLIN Properties
"Catalonia is positioned as a strategic technology hub and combines a strong foundation of renewable energy with an extremely efficient, CO2-neutral base with nuclear infrastructure".
Microsoft Announces Layoffs Affecting Software Engineers
Microsoft has announced a significant reduction in its workforce, cutting approximately 7,000 employees (about 3% of its global staff). Software engineering positions have been particularly impacted, with reports indicating they constitute over 40% of the roughly 2,000 positions eliminated in Washington state alone. The layoffs come as Microsoft continues to invest heavily in artificial intelligence infrastructure and capabilities, forcing a realignment of its workforce as the company adapts to the ongoing AI transformation of the tech industry.
GETTY IMAGES
"The layoffs, reported on May 13, 2025, are the biggest cuts at the company since the elimination of 10,000 roles in 2023".
AI Development & Product Updates
What to Expect at Google I/O 2025
Google I/O 2025 promises to be a landmark event for AI development, with several major announcements anticipated. Attendees can expect significant updates to the company's Gemini AI models, including potential releases in the areas of AI agents, Android XR, and expanded memory capabilities for personalized AI. The conference is likely to showcase Google's strategy for integrating AI more deeply across its ecosystem, with special attention to the newly developed Android XR platform that could power smart glasses and other extended reality devices.
"Much like last year, we may see several new AI announcements related to Gemini, AI agents, Android XR, and more".
Cursor AI Releases Major Update
Cursor AI, the popular AI-powered code editor, has released significant updates to its platform, enhancing its capabilities for developers. Recent changes include integration with OpenAI's o3 and o4-mini models for improved coding assistance, offering GPT-4.1 access to all users for a limited time, and introducing new features like error auto-fixing, navbar enhancements, and code previews. The platform has also updated its pricing structure with a simpler token-based system for its Max Mode, which provides access to the most powerful AI features.
Image by Jim Clyde Monge
"Cursor, the popular AI vibe coding platform, just dropped a series of new features in its v0.50 update".
GenerativeAI.pub: Cursor 0.5 Just Dropped— Here's All The Features You Need To Know
Manus AI Updates Platform and Adds Image Generation
Manus AI has made significant updates to its platform, removing the invite-only system to allow public registration and introducing advanced image generation capabilities. The company now offers free daily tasks worth 300 credits and a one-time bonus of 1,000 credits for new users. What sets Manus AI's image generation apart is its integration within a broader task execution process - rather than simply generating images, it understands user intent, plans solutions, and combines multiple tools to accomplish complex goals, positioning image creation as just one step in achieving larger user objectives.
Introducing Manus image generation.
Manus doesn’t just generate images. It understands your intent, plans a solution, and knows how to effectively use image generation along with other tools to accomplish your task.
— ManusAI (@ManusAI_HQ)
4:11 PM • May 15, 2025
"Manus doesn't just generate images. It understands your intent, plans a solution, and knows how to effectively use image generation along with other tools to accomplish your task".
Conclusion
This week's developments highlight the dual nature of AI's evolution - while capabilities continue to expand with more autonomous agents, improved code generation, and creative AI applications, we're also seeing more realistic assessments of current limitations. The infrastructure supporting these advances is growing more robust, as evidenced by CoreWeave's European expansion, but corporate restructuring at Microsoft reminds us of the significant costs associated with AI transformation. As autonomous AI agents become more mainstream, we're likely to see continued refinement of the human-AI collaboration model rather than complete replacement, with humans maintaining oversight of complex decision-making while delegating routine tasks to increasingly capable AI systems.
The Author
 Luis Poveda’s AI Newsletter | Luis Poveda is a technology optimist and passionate innovator, constantly exploring and researching the latest trends. Based in Barcelona, he is currently focused on AI and developing a modern AI-driven IT network observability tool. He is also the creator and maintainer of Luis Poveda's AI Newsletter, where he curates and shares key insights on the evolving AI landscape. |
Start learning AI in 2025
Keeping up with AI is hard – we get it!
That’s why over 1M professionals read Superhuman AI to stay ahead.
Get daily AI news, tools, and tutorials
Learn new AI skills you can use at work in 3 mins a day
Become 10X more productive