Luis Poveda's AI Newsletter: May 26, 2025

In partnership with

Start learning AI in 2025

Everyone talks about AI, but no one has the time to learn it. So, we found the easiest way to learn AI in as little time as possible: The Rundown AI.

It's a free AI newsletter that keeps you up-to-date on the latest AI news, and teaches you how to apply it in just 5 minutes a day.

Plus, complete the quiz after signing up and they’ll recommend the best AI tools, guides, and courses – tailored to your needs.

Sign up to start learning.

This week marked a pivotal moment in AI development as major players unveiled transformative advances that bring artificial intelligence closer to human-like capabilities. Google dominated headlines with its I/O announcements, introducing Deep Think reasoning for Gemini 2.5 Pro and launching the enhanced Gemini 2.5 Flash alongside the creative powerhouse Veo 3 video generator. Meanwhile, Anthropic claimed the coding crown with Claude 4's autonomous development capabilities, Mistral democratized programming assistance with its open-source Devstral model, and OpenAI made its boldest hardware bet yet by acquiring Jony Ive's design firm for $5 billion, signaling a new era where AI software meets world-class industrial design.

Google's Gemini 2.5 Gets Deep Think Reasoning Powers to Challenge OpenAI's o3

Google unveiled Deep Think, an enhanced reasoning mode for Gemini 2.5 Pro that allows the model to consider multiple answers before responding, similar to OpenAI's o1 and o3 reasoning models. The technology uses parallel processing techniques to boost performance on complex tasks, with Gemini 2.5 Pro Deep Think achieving top scores on LiveCodeBench coding evaluations and surpassing OpenAI's o3 on MMMU perception and reasoning tests. Currently available to trusted testers via the Gemini API, the feature represents Google's strategic push to compete with OpenAI's reasoning-focused models while the company conducts additional safety evaluations before wider deployment.

• TechCrunch: 'Deep Think' boosts the performance of Google's flagship Google Gemini AI model

• WinBuzzer: Google Unveils Gemini 2.5 Pro 'Deep Think' for Advanced AI Reasoning, Improves 2.5 Flash Model

Gemini 2.5 Flash Emerges as Google's High-Performance Workhorse Model

Google launched an updated Gemini 2.5 Flash model that delivers major improvements in reasoning capabilities while maintaining the speed and cost efficiency that made its predecessor popular among developers. The new version offers enhanced performance on coding, multimodal tasks, reasoning, and long-context processing, with Google positioning it on the "pareto frontier" for performance-to-cost ratio. Available in preview through Google AI Studio, Vertex AI, and the Gemini app, the model supports up to 1 million input tokens and integrates with new features like Canvas for interactive document and code refinement, with general availability planned for developers and enterprises in early June.

• Google Blog: Gemini 2.5 Flash is now in preview

• Google Cloud: Gemini 2.5 on Vertex AI: Pro, Flash & Model Optimizer Live

Anthropic Claims Coding Crown with Claude 4's Autonomous Development Capabilities

Anthropic released Claude 4 Opus and Claude 4 Sonnet, with the company boldly claiming the Opus model as "the world's best coding model" capable of autonomous operation for hours on complex development tasks. The new models introduce "extended thinking with tool use," a beta feature that combines reasoning with external tool access including web search, similar to OpenAI's o3 capabilities. Claude 4 Opus demonstrated remarkable endurance by refactoring code continuously for seven hours, while both models show significant improvements in coding abilities, instruction adherence, and agentic AI applications, targeting the growing demand for AI systems that can work independently on software engineering challenges.

• Ars Technica: New Claude 4 AI model refactored code for 7 hours straight

• Gadgets 360: Anthropic Releases Claude 4 Series AI Models With Improved Coding Capability and Tool Use

Google's Veo 3 Revolutionizes AI Video Creation with Native Audio Generation

Google announced Veo 3, its latest AI video generation model that introduces native audio capabilities, allowing creators to generate videos with integrated ambient sounds, background music, and dialogue rather than silent clips. The model features improved text-to-video prompt recognition and works alongside the new Flow AI filmmaking tool, which combines Veo with Imagen and Lyria models to create cinematic scenes from simple text descriptions. Available to Google AI Pro and Ultra subscribers, Veo 3 represents a significant advancement in creative AI tools, competing directly with OpenAI's Sora while offering the unique advantage of synchronized audio-visual generation that could transform content creation workflows.

• Gadgets 360: Google Unveils Video-Generation Model Veo 3 and Image Generation Model Imagen 4 at Google I/O 2025

• Financial Express: Google I/O 2025: Google announces Veo 3, a new AI video generator that creates clips with realistic audio

Mistral Democratizes AI Coding with Open-Source Devstral Model

French AI startup Mistral launched Devstral Small 24B, an open-source coding model developed in partnership with All Hands AI that achieves a record 46.8% score on the SWE-Bench Verified benchmark. Released under the Apache 2.0 license, Devstral outperforms other open models including Google's Gemma 3 27B and DeepSeek's V3, while being compact enough to run on laptops with just 24 billion parameters. The model is designed as an "agentic" coding assistant for integration into development environments, IDEs, and plugins, representing Mistral's strategic focus on specialized, efficient models that can be deployed locally, replacing their previous Codestral model which had commercial use restrictions.

• TechCrunch: Mistral's new Devstral AI model was designed for coding

• The Decoder: Mistral launches Devstral Small 24B, a new open-source LLM for coding

OpenAI Bets $5 Billion on Jony Ive's Design Vision for AI Hardware Future

OpenAI completed a $5 billion acquisition of Io, the hardware startup co-founded by former Apple design chief Jony Ive and Sam Altman, merging the renowned designer's creative collective with OpenAI's AI expertise to develop next-generation consumer devices. The collaboration, which began two years ago, brings together Ive's team including former Apple executives Tang Tan, Scott Cannon, and Evans Hankey, who will handle design and engineering while OpenAI provides AI capabilities. The acquisition represents OpenAI's most ambitious hardware venture, aiming to create AI-powered devices that could revolutionize human-computer interaction, though the companies remain secretive about specific product details, promising to share their work in 2026.

• WIRED: OpenAI's Big Bet That Jony Ive Can Make AI Hardware Work

• MacRumors: OpenAI Buys Jony Ive's AI Startup to 'Completely Reimagine What It Means to Interact with Computing'

We’re so early

This week brought a slew of truly impressive announcements from Google. Sundar Pichai, who not long ago may have felt the search empire slipping, now exudes confidence, firmly asserting Google’s leadership across multiple domains. Their breakthroughs in large language models are powering not only consumer applications but also cutting-edge research in science and mathematics. In video generation, too, Google is setting industry benchmarks. Even Claude, long hailed as the coding assistant leader, faces stiff competition as Google steadily captures more of the developer mindshare. One can only hope Tim Cook finds his own “Sundar moment” soon.

Yet the most unexpectedly human announcement came from OpenAI’s acquisition of io, the design studio founded by legendary Apple designer Jony Ive. It’s ironic that news of AI-centric hardware felt so personal, a reminder of how extraordinary talent can craft beautiful, world-changing products that can make a ding in the universe. If you haven’t seen the reveal video, it’s well worth your time.

This week's developments signal a maturing AI landscape where raw computational power increasingly meets sophisticated reasoning, creative capabilities, and thoughtful design. Google's comprehensive I/O announcements demonstrate the tech giant's commitment to competing across the entire AI stack, from reasoning models to creative tools. Anthropic's bold claims about Claude 4's coding supremacy challenge the established order, while Mistral's open-source approach ensures democratized access to advanced capabilities. Most intriguingly, OpenAI's massive investment in Jony Ive's design expertise suggests the industry recognizes that AI's next breakthrough may not come from better algorithms alone, but from reimagining how humans interact with intelligent systems through beautifully designed hardware interfaces.

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

Sign up and start mastering AI