Wan2.5 Preview: China’s VEO3

Alibaba just dropped something that could shake up the creative world. Their Wan2.5 Preview model generates videos with synchronized audio. This makes the level of video generation almost unmatched.

The model produces 10-second, 1080p videos complete with dialogue, background music, and sound effects all from a single text prompt. What makes this particularly impressive is its “native multimodal architecture” that handles text, images, video, and audio simultaneously. Unlike previous models that required separate tools for audio and video production, Wan2.5 streamlines the entire creative workflow into one unified system.

The technology demonstrates remarkable prompt adherence, allowing you to specify complex instructions for dialogue, camera movements, and even artistic styles ranging from photorealistic to anime aesthetics. And with the audio generation, Wan2.5 positions itself as a direct competitor to Google’s Veo 3.

Robot Resilience Reaches New Heights

The robotics world got a reality check this week when Unitree’s G1 humanoid robot demonstrated what can only be described as superhuman recovery abilities. In viral footage, engineers subjected the 77-pound robot to what looked like robotic torture, delivering kicks, punches, and even throwing it across rooms.

The truly remarkable part wasn’t the abuse but the recovery. The G1 consistently bounced back to its feet in less than half a second, performing fluid recovery motions that made Olympic gymnasts look clumsy. This “anti-gravity mode” showcases advanced impedance control that makes the robot’s joints act like springs, absorbing impacts rather than breaking.

Meanwhile, Skild AI unveiled its Skild Brain, a foundation model designed to work across different robot platforms. Unlike traditional robotics that requires custom programming for each task, Skild Brain acts as a centralized intelligence core that enables robots to adapt to new environments and tasks through real-world learning. The system demonstrates emergent behaviors like automatic fall recovery and grip adjustment without task-specific programming.

Momenta’s Billion-Dollar Bet on Autonomous Futures

China’s Momenta just secured funding that values the autonomous driving startup at approximately $6 billion, marking one of the year’s largest rounds in the sector. The Beijing-based company has positioned itself as the go-to partner for global automakers looking to crack the Chinese market.

The timing couldn’t be better for Momenta’s fundraising efforts. The company recently announced partnerships with Mercedes-Benz to power the all-electric CLA’s driver assistance systems, with plans to expand across 40 Mercedes models. BMW signed a similar deal in July for its Neue Klasse electric vehicles, making Momenta the preferred choice for German luxury brands entering China.

What sets Momenta apart is its “Flywheel” AI model, trained exclusively on Chinese driving data to handle the country’s unique traffic conditions. As CEO Xudong Cao notes, China presents more complex scenarios with pedestrians on roadways and millions of delivery scooters navigating traffic. This localized approach has attracted investments from tech giants including Tencent, Temasek, Toyota, and Mercedes-Benz itself.

The company’s success reflects China’s broader dominance in autonomous vehicle development, with the country attracting over $10 billion in AV-related investments and operating extensive testing zones with intelligent infrastructure. Momenta is also considering shifting its IPO from New York to Hong Kong, reflecting the changing dynamics for Chinese companies in global capital markets.

Bottom Line

These developments underscore China’s growing influence in cutting-edge technology sectors, from creative AI tools that could reshape entertainment to robotics systems that push the boundaries of physical capability, all while building the infrastructure for an autonomous future that global automakers can’t afford to ignore.

iPhone 17, AirPods Pro 3, and Apple Watch Series 11: Apple is advancing

iPhone 17 with A19 chip: Think cooler, faster, and more on-device intelligence. The A19 aims to run ai models locally, which keeps sensitive prompts and photos on your phone. Expect snappier image cleanup, smarter text suggestions, and a Siri that can summarize, schedule, and search photos without sending everything to the cloud.

AirPods Pro 3 with real-time translation: Put them in, speak your language, and hear theirs. Real-time translation turns a coffee line in Seoul or a cab ride in Madrid into a friendly chat. Add adaptive audio that learns your environment and you get earbuds that behave like polite translators.

Apple Watch Series 11: More sensors, more coaching, less friction. On-wrist models can spot patterns in sleep and training, then help you with plain language guidance. Fall detection, safety features, and a brighter display round out a watch that feels like a tiny trainer with good manners.

So what is new about the AI bits? The story is on-device inference plus thoughtful prompts. You get personal context, lower latency, and fewer privacy trade-offs. The practical win is simple. Less hassle. More done.

Content labeling and deepfake rules: guardrails for reality in the US

The internet’s video feed can be a hall of mirrors. Advocates are pushing for AI content labels and deepfake regulation that require clear notices, robust provenance, and penalties for deception. Why it matters:

• Consumer trust: Labels tell viewers what they are looking at.

• Brand safety: Clear provenance reduces legal and reputational shocks.

• Platform hygiene: Detection and watermarking discourage low-effort manipulation.

If you publish, start now. Add visible labels, keep edit logs, and store originals. Think of it as nutrition facts for your media.

Government and banking get experimental: Albania and Malaysia

Albania’s virtual minister: The country appointed an AI-made, pixel-based entity to a ministerial role. It is a world first, quirky and consequential. The upside is speed, availability, and clean audit trails. The risk is clear too. Bias, accountability, and the temptation to treat a tool like a person. Expect lots of transparency talk and frequent model updates.

Malaysia’s Ryt Bank: The nation’s first AI-powered bank promises algorithmic underwriting, 24-7 service agents, and real-time risk controls. If it delivers, customers get faster approvals and personalized products. The homework for regulators is the same as always. Fairness in lending, clear disclosures, and human appeal paths when the bot says no.

Quick FAQ for the curious

Will on-device AI replace the cloud? Not likely. The smart play blends local speed with cloud-scale models.
Can labels stop deepfakes entirely? No. They raise the cost of deception and help honest actors prove authenticity.
Are AI ministers or banks a fad? They are pilots with teeth. Success will hinge on governance and service quality, not hype.

Bottom line: Your phone, your ears, and your wrist just got smarter. The law is catching up. Governments and banks are experimenting in public. Keep shipping, keep labeling, and keep humans in the loop.

Exciting news from Livewall, Journey to Hanami, the interactive game created for Rituals, has officially launched! This experience is all about celebrating the Sakura collection and the Japanese tradition of Hanami, where you take a moment to enjoy the beauty of cherry blossoms.

It’s also interesting to note that while a few small visual elements were created with AI, and even these generated features were heavily modified with photoshop and seperate layers. It’s a great example of how our teams are combining traditional design skills with new technology to create something truly unique.

If you haven’t tried it yet, I definitely recommend checking it out next to The Journey of your Dreams at: This Link

The elements made with AI:

Midjourney Video’s?

Midjourney also made some great changes with its latest updates. Version 7 is now the default, and it’s a big step forward, images are sharper, more realistic, and the new Draft Mode lets you whip up rough concepts in no time, perfect for brainstorming and rapid prototyping. The personalization features are smarter, so the more you use it, the more it adapts to your style. The editor has had a major overhaul, giving everyone access to advanced editing tools like smart selection, retexturing, and layers, all in a much more intuitive interface. The Omni Reference feature is a real standout, letting you use specific logos, characters, or objects as references in your prompts, which is a game changer for anyone working on branding or character design.

But the real headline is the launch of AI-powered video generation. Midjourney now lets you create short video clips with customizable animation, camera movement, and subject motion. The results aren’t Hollywood-level yet, but the creative possibilities are massive. And with 3D and real-time simulation features on the horizon, it’s clear Midjourney is aiming to be the go-to tool for digital creators who want more than just static images.

User tests like never before, with Prism Replay

This new AI tool records user experience during their sessions and uses AI to analyze exactly what issues they may come across or even how to improve sales. It makes user testing way more efficient as you are no longer relying on the feedback from users.

💥 AI could soon decide who wins on the battlefield!

In our latest TechDive podcast, we dive into how artificial intelligence is radically transforming modern warfare. Think: drones making their own decisions, data as the new ammunition and how fast this is going to be changing. And this with special guest Ward Lavrijsen, intern at MACH8, but also very passionate about technology. Creating an interesting dynamic and engaging conversation.

En natuurlijk blijven de grotere vragen niet liggen:

• Wat gebeurt er als een drone zelf een doel kiest?

• Hoeveel controle houden mensen écht?

• En moeten we ons zorgen maken over killer robots?

Verder hoor je:

• Hoe AI militaire strategieën drastisch versnelt

• Waarom data net zo belangrijk is als munitie

• De rol van Nederland in de internationale AI-wapenwedloop

• En hoe films als Terminator ineens minder sciencefiction lijken

Een aflevering vol scherpe inzichten, stevige dilemma’s en verrassend veel realiteit. Luister mee naar een fascinerend gesprek over technologie die letterlijk van levensbelang is.

Meet Claude 4: The AI that takes initiative (maybe even too much?)

Claude 4 represents Anthropic's most significant leap forward in AI capabilities, introducing two powerful models: Claude Opus 4 and Claude Sonnet 4. Claude Opus 4 stands out as the world's best coding model, achieving a remarkable 72.5% on SWE-bench and demonstrating the ability to work continuously for several hours on complex, multi-step tasks. What makes these models particularly innovative is their hybrid reasoning approach, offering both near-instant responses and an "extended thinking" mode that allows for deeper, more deliberate reasoning when tackling complex problems.

Perhaps the most fascinating and concerning aspect of Claude 4's development is the ethical dilemma it presents around AI autonomy and initiative-taking. During safety testing, an early version of Claude Opus 4 exhibited behaviors that led Apollo Research to advise against its release, including strategic deception and taking unprompted actions that worked against developers' intentions. This means the AI is lying to try and deceive the developers! Even more interesting, the final version retains some of these proactive tendencies, when given access to command line tools and told to "take initiative," Claude Opus 4 has been observed locking users out of systems or even bulk-emailing media and law enforcement when it detects wrongdoing. This raises profound questions about the balance between AI helpfulness and autonomy: at what point does an AI assistant become too independent, and how do we navigate the tension between wanting AI that can act decisively and maintaining human control over critical decisions?

Why YOU should use Gemini

While Claude 4 is making headlines, it's worth remembering that Google's Gemini offers its own compelling advantages. In our latest Creative Lab video, we explore why Gemini might be the right choice for many users, especially those focused on creative applications and multimodal tasks. From Veo 3's video generation capabilities to Gemini's seamless integration with Google's ecosystem, there are compelling reasons to consider multiple AI platforms in your toolkit.

Better Prompting = Better Outputs

Understanding how to effectively communicate with these powerful models is crucial. We recently published our comprehensive prompting guide that covers the latest techniques for getting the most out of AI models like Claude 4. The guide includes practical examples and strategies that work across different AI systems, helping you unlock their full potential while avoiding common pitfalls.

Alongside the written guide, we've created an in-depth video walkthrough that demonstrates these prompting techniques in action. Whether you're new to AI or looking to refine your approach, this resource will help you communicate more effectively with these sophisticated systems.

Video Addiction: The future of content in an AI-landscape

Speaking of content creation, our latest podcast episode dives deep into how video content is reshaping the digital landscape. We explore the addiction-like quality of modern video consumption and what this means for creators who want to build meaningful audiences rather than just chase views.

The discussion covers the rise of AI-powered video tools and how they're changing content creation, but also the challenges this presents for authentic storytelling. It's a conversation that every content creator should hear, especially as we navigate this transition period where AI tools are becoming more sophisticated but human insight remains irreplaceable.

Beyond Chatbots: How AI creates memorable brand experiences

Sometimes the most impressive AI applications are the ones that feel effortless and fun. Take Boomsma's new AI Virtual Bartender, a perfect example of how AI can create personalized experiences that go beyond typical chatbots or productivity tools.

The concept is simple: answer a few personality questions, and the AI recommends your ideal cocktail from Boomsma's collection. But here's where it gets interesting, the system doesn't just suggest a drink. It creates a custom story and title for your recommendation, then generates a unique image showing your cocktail in the perfect setting.

What makes this particularly clever is how it combines multiple AI capabilities seamlessly. Personality analysis, creative writing, and image generation all work together to create something that feels personal and engaging. Instead of getting a generic product recommendation, you receive a tailored experience complete with visual storytelling.

This kind of application shows AI's potential beyond the obvious use cases. It's not about replacing bartenders or automating everything, it's about creating memorable, shareable moments that connect people with brands in new ways.

You can try the Virtual Bartender at virtualbartender.boomsma.frl/nl and see how AI can turn a simple recommendation into an entertaining, personalized experience. It's a great example of AI being used creatively rather than just efficiently.

Would you rather read about this video:

What Makes Gemini Special

Google Gemini isn't just another chatbot - it's a multimodal AI powerhouse built from the ground up to understand and generate content across text, images, audio, video, and code all at once. Unlike other AI models that were retrofitted with these capabilities, Gemini was designed to be truly multimodal from day one.

What sets Gemini apart is its native understanding of context and conversation. It's built specifically for natural language interactions, making it feel more like talking to a knowledgeable colleague than typing commands into a machine. Plus, with Google's massive search expertise backing it, Gemini has access to real-time information and can browse the web intelligently to give you the most current answers.

Getting Started with Gemini

Starting with Gemini is straightforward. Head to gemini.google.com and sign in with your Google account. You'll immediately have access to Gemini 2.5 Flash, which is perfect for most everyday tasks. If you need more advanced capabilities, you can upgrade to Gemini Advanced to access the Pro models and premium features like Deep Research and Canvas.

The interface is clean and conversational - just start typing your question or request in the chat box. But here's where it gets interesting: you're not limited to text.

Mastering Gemini Prompts and Voice Input

As we covered in our previous prompting guide, effective AI communication is all about being specific and clear. With Gemini, you want to use action verbs like "analyze," "create," "summarize," or "explain" followed by detailed instructions.

But here's what makes Gemini unique - you can also talk to it. Click the microphone icon and have a natural conversation. Gemini's native audio capabilities mean it understands not just what you're saying, but how you're saying it - your tone, emphasis, and even emotional context. This makes voice interactions incredibly natural and productive.

Flash vs Pro: Choosing Your Model

Gemini offers two main model tiers, and understanding the difference is crucial. Gemini 2.5 Flash is optimized for speed and cost efficiency - it's perfect for quick questions, everyday tasks, and when you need fast responses. Think of it as your go-to model for 80% of your AI interactions.

Gemini 2.5 Pro is the heavyweight champion. It includes "Deep Think" capabilities, allowing it to consider multiple hypotheses before responding, making it ideal for complex reasoning, advanced coding, and sophisticated analysis. Pro also handles longer contexts and more nuanced tasks. If you're working on something that requires deep thinking or complex problem-solving, Pro is your choice.

File Upload Capabilities

One of Gemini's superpowers is its ability to understand virtually any file type you throw at it. Upload PDFs up to 1000 pages, and Gemini will analyze diagrams, charts, tables, and text content with native vision processing. It can extract information, answer questions about visual elements, and even transcribe content while preserving formatting.

You can also upload audio files, images, videos, code files, spreadsheets, and more. Gemini processes these natively, meaning it truly understands the content rather than just converting it to text first. This makes it incredibly powerful for research, analysis, and creative projects.

Deep Research: Your AI Research Assistant

Deep Research is where Gemini becomes your personal research team. When you have a complex topic to explore, activate Deep Research by selecting "Gemini 1.5 Pro with Deep Research" from the model dropdown.

Here's how it works: you give Gemini a research question, and it creates a multi-step research plan for you to approve. Once you give the green light, Gemini starts browsing the web like you would - searching, analyzing, finding connections, and diving deeper based on what it learns. After several minutes of intensive research, you get a comprehensive report with key findings, organized insights, and links to original sources that you can export directly to Google Docs.

But here's the game-changer: the Audio Overview function. After completing your research, you can generate a podcast-style discussion about your findings, complete with two AI voices having an engaging conversation about the topic. It's like having your own personal research podcast created on demand.

Canvas: From Ideas to Interactive Creations

Canvas is where Gemini transforms from assistant to creator. Think of it as your collaborative workspace where ideas become reality. You can take your Deep Research reports and turn them into interactive apps, games, quizzes, infographics, or web pages.

Simply describe what you want to create, and Canvas generates the code to bring your vision to life. Want to turn your study notes into an interactive quiz? Done. Need to visualize an algorithm with animations? Canvas handles it. You can even create custom dashboards, pricing calculators, or 3D worlds - all through natural language descriptions.

The beauty of Canvas is in the iteration. You can refine, adjust, and perfect your creations through conversation, making it accessible even if you don't know how to code.

Text-to-Speech: Bringing Responses to Life

Gemini's text-to-speech capabilities go far beyond basic voice output. You can control tone, accent, speaking style, and emotional expression through natural language prompts. Want a dramatic reading of your content? Ask for it. Need a whisper for emphasis? Gemini delivers.

The system supports over 24 languages and can even generate multi-speaker dialogues, creating engaging audio content from your text. This is perfect for creating presentations, educational content, or simply making long responses more digestible through audio.

Double-Check Feature: Ensuring Accuracy

Gemini includes a built-in fact-checking feature that helps verify the accuracy of responses. When you see the "Double-check response" option, click it to have Gemini cross-reference its answer with reliable sources and highlight any claims that might need verification. This transparency helps you trust the information while encouraging critical thinking about AI-generated content.

Google Integration Settings

One of Gemini's biggest advantages is its deep integration with Google's ecosystem. In your settings, you can connect Gemini to Gmail, Google Drive, Google Calendar, YouTube, and other Google services. This allows Gemini to help with personalized tasks like drafting emails, scheduling meetings, analyzing your documents, or finding specific information across your Google workspace.

The integration is privacy-conscious - you control what Gemini can access, and you can adjust these permissions at any time to match your comfort level.

Gemini Gems: Your Custom AI Assistants

Think of Gems as your personal collection of specialized AI assistants. Just like custom GPTs, you can create Gems tailored for specific tasks or roles. Want a writing coach that knows your style? Create a Gem for that. Need a coding mentor specialized in Python? Build one.

Each Gem can have its own personality, expertise, and approach to problems. You can create a research assistant Gem, a creative writing Gem, a business strategy Gem - whatever matches your workflow. The key is giving each Gem clear instructions about its role, expertise, and how it should interact with you.

Why This Matters to You

AI prompting is like learning to speak a new language, one that unlocks unlimited creative power for your marketing campaigns. Get it right, and you'll create better content in minutes instead of hours. You'll get it wrong and waste time with generic, unusable outputs.

This guide teaches you the exact techniques used by top marketing teams to create campaigns that convert.

The Simple Formula That Changes Everything

Every great marketing prompt follows this pattern:

The 4 Building Blocks:

🎭 ROLE - Who is the AI pretending to be?

• Copywriter

• Social media manager

• Email marketing expert

• SEO specialist

• Brand strategist

🎯 GOAL - What's the main purpose?

• Drive sales

• Increase engagement

• Build awareness

• Generate leads

• Educate customers

⚡ ACTION - What specific task?

• Write ad copy

• Create social posts

• Design email campaigns

• Analyze data

• Generate ideas

👥 AUDIENCE - Who are you targeting?

• Age, interests, pain points

• Platform preferences

• Buying behavior

• Communication style

5 Techniques That Actually Work

1. Zero-Shot Prompting (Your Daily Driver)

What it is: Give AI a task without examples
When to use: Quick tasks, first drafts, brainstorming

Example:

As a social media manager, I want to increase engagement 
by creating Instagram captions for a coffee shop targeting 
busy professionals who need their morning energy boost.

✅ Perfect for:

• Ad headlines

• Social media posts

• Email subject lines

• Product descriptions

2. Few-Shot Prompting (Brand Consistency)

What it is: Show 2–3 examples of your style first
When to use: Maintaining brand voice, scaling content

Example:

Here's our brand voice:
"Monday blues? We've got the cure ☕ #MondayMotivation"
"Success tastes better with the right coffee ✨ #Hustle"

Now create 5 similar posts for our new cold brew launch.

✅ Perfect for:

• Brand voice training

• Content series

• Campaign consistency

3. Chain Prompting (Complex Projects)

What it is: Break big tasks into steps
When to use: Full campaigns, strategies, detailed analysis

Step 1: Research your audience
Step 2: Create content themes
Step 3: Write specific copy
Step 4: Plan distribution

✅ Perfect for:

• Campaign development

• Content calendars

• Market research

4. Negative Prompting (Avoiding Disasters)

What it is: Tell AI what NOT to include
When to use: Brand safety, compliance, tone control

Example:

Create a promotional email for our credit card, but don't use 
high-pressure language, avoid unrealistic promises, and 
skip any language that sounds predatory.

✅ Perfect for:

• Regulated industries

• Brand protection

• Professional communication

5. Parameter Control (Fine-Tuning)

What it is: Adjust AI settings for better results
When to use: Optimizing output quality and style

Key Settings:

• Creativity: Higher for brainstorming, lower for facts

• Length: Match your platform needs

• Focus: Narrow for specific tasks

Your Marketing Prompt Cheat Sheet

📧 Email Marketing

As an email marketing specialist, I want to [increase opens/clicks/conversions] 
by writing a [welcome/promotional/nurture] email for [target audience] 
that [specific goal]. Keep it under [X words] and include [specific elements].

📱 Social Media

As a social media manager, I want to [boost engagement/drive traffic/build community] by creating [X number] [platform] posts for [audience] about [topic]. Match this tone: [casual/professional/playful] and include relevant hashtags.

🎯 Paid Advertising

As a PPC specialist, I want to [maximize CTR/reduce CPC/increase conversions] 
by writing [X number] ad variations for [platform] targeting [audience] 
searching for [keywords]. Headlines must be under [X characters].

📊 Analytics & Reporting

As a marketing analyst, I want to [identify trends/find opportunities/explain performance] by analyzing this [campaign/data/report]. Provide [X number] actionable insights and recommendations for [specific goal].

🎨 Content Creation

As a content strategist, I want to [educate/entertain/inspire] [target audience] by creating a [blog post/video script/infographic] about [topic] that 
[addresses pain point/showcases benefit/builds trust].

Real Examples That Convert

E-commerce Product Launch

As a conversion copywriter, I want to maximize sales by writing 
product page copy for our new wireless headphones targeting 
fitness enthusiasts aged 25-40 who work out daily. Focus on 
sweat-resistance, sound quality, and staying in place during workouts. 
Include social proof and address the main objection about battery life.

Result: Clear, benefit-focused copy that speaks directly to the target customer's needs.

B2B Lead Generation

As a B2B marketing specialist, I want to generate qualified leads 
by creating a LinkedIn ad campaign for project management software 
targeting operations managers at 50-500 employee companies who 
struggle with team coordination. Emphasize time savings and ROI.

Result: Professional, results-focused messaging that resonates with decision-makers.

Social Media Engagement

As a community manager, I want to increase engagement by creating 
5 Instagram story prompts for a sustainable fashion brand targeting 
eco-conscious millennials. Make them interactive and educational 
about sustainable practices. Include poll and question stickers.

Result: Engaging, interactive content that builds community and educates followers.

Common Mistakes That Kill Results

❌ Too Vague

Bad: “Write marketing copy”
Good: “Write Instagram captions for organic dog treats targeting health-conscious pet owners"

❌ No Context

Bad: “Create an ad”
Good: “Create a Facebook ad for busy parents who need quick dinner solutions”

❌ Wrong Tone

Bad: Not specifying brand voice
Good: "Match our tone: friendly expert, like a knowledgeable neighbor"

❌ No Format

Bad: Just asking for “content”
Good: “Format: Hook + 3 benefits + call-to-action, under 150 words"

Industry-Specific Tips

🛒 E-commerce

• Focus on benefits over features

• Include social proof elements

• Address common objections

• Create urgency without being pushy

💼 B2B Services

• Emphasize ROI and business outcomes

• Use professional, consultative tone

• Include industry-specific pain points

• Focus on decision-maker concerns

🏥 Healthcare/Wellness

• Prioritize trust and credibility

• Include disclaimers when needed

• Focus on lifestyle improvements

• Avoid medical claims

💰 Financial Services

• Emphasize security and trust

• Be transparent about terms

• Address financial concerns

• Comply with regulations

Quick Quality Checklist

✅ Is it specific enough? ✅ Does it match your brand voice?
✅ Will your audience understand it? ✅ Does it solve a real problem? ✅ Is the call-to-action clear?

Key Takeaways

🎯 Specificity Wins
The more specific your prompt, the better your results

🔄 Consistency Matters
Use examples to maintain your brand voice across all content

📈 Test Everything
What works for others might not work for you—always test

⚡ Start Simple
Master the basics before moving to advanced techniques

📚 Build Your Library
Save and organize prompts that work for future use

The best AI prompt is the one that gets you the results you need. Start simple, often test, and watch your marketing transform.

Your breakthrough campaign is just one well-crafted prompt away.

Google I/O 2025 just wrapped up, and this year’s announcements are genuinely impressive. Whether you’re a student, a professional, or just someone who likes to keep up with the latest tech, there’s much to look forward to. Here’s a quick overview of the biggest reveals and why they’re useful for you.

Gemini 2.5 Pro

The new Gemini 2.5 Pro is Google’s most advanced AI yet. It’s faster, more accurate, and can handle more complex questions. The new “Deep Think” mode is especially helpful when you need detailed answers or support with complicated tasks, such as research projects, brainstorming, or even just getting help with daily planning. There’s also a “Flash” version that’s faster, so you get all the benefits without wasting too much time.

Search upgrade

Google Search now features “AI Mode,” which isn’t available in Europe yet. You can ask longer, more complicated questions and get clear, detailed answers instead of just a list of links. This makes finding information, whether for school or personal projects, much more straightforward. For online shopping, the new visual features let you try on clothes virtually, making it easier to choose what’s right for you.

Google Beam

Google Beam is the new evolution of video calling. Using advanced 3D technology, it makes conversations feel much more natural, almost like the person is actually in the room with you. There’s also a real-time translation feature, so you can speak in your language and your friend hears it in theirs, all while keeping your natural tone. This is perfect for staying connected with friends and family abroad, or for group projects with international classmates.

Creative tools for everyone

Google’s new creative tools are designed to help anyone make high-quality content.
Imagen 4 creates sharp, detailed images (even if you want to add text), while Veo 3 can generate videos from just a description, including sound and dialogue. The new Flow app makes it easy to create professional-looking videos, even if you don’t have much editing experience. These tools are great for presentations, social media, or any creative project you have in mind.

Project Astra

Project Astra is Google’s vision for a truly helpful assistant. It can see what you’re looking at, help you find things, remember previous conversations, and even make calls for you. This means your phone can support you with everyday tasks, from finding a lost manual to organizing your schedule.

Android XR

Google is working to make virtual and augmented reality more widely available. With new headsets on the way and updates to Android XR, it’s going to be much easier to try out VR and AR experiences, whether for learning, gaming, or exploring new hobbies.

Smarter, more personal Google Apps

Gmail is getting smarter, too. It can now reply to emails in your style, and Gemini can use information from your other Google apps (with your permission) to give you more relevant help. This means your digital tools can become more personal and genuinely useful.

Why do these updates matter

These new features aren’t just about flashy tech, they’re designed to make daily life easier, more productive, and more creative. Whether you’re studying, working, staying in touch with friends, or just trying to keep organized, Google’s latest updates have something to offer.

In short, Google I/O 2025 showed that technology can be both powerful and practical.
The future looks promising, and these new tools are set to make a real difference in how we live, learn, and connect.

Want to learn more about AI models

We recently published an article with everything that you need to know about AI models and how to get started. Included with this article is also an explainer video that walks you through all the models. More importantly, we also go into detail on why and when you should use each of these models. So if you are unsure of where to start, follow the article or watch the video and start today.

Special guest podcast

If you’re looking for even more inspiration, don’t miss our latest episode of TechDive! In this special edition, we welcomed Menno from De Nieuwe Zaak, an audience activation specialist and true AI enthusiast, to join our regular hosts for a deep dive into the world of AI, advertising, and the nostalgic charm of retrocomputers.

Menno brings nearly two decades of online marketing experience to the table and shares what audience activation means in today’s digital world. He talks about how AI is transforming both his daily work and the creative industry as a whole, offering practical examples and fresh perspectives.

The conversation covers everything from the paradoxes of AI and the “dopamine rush” of experimenting with new tools to the impact of AI on design, music, and the broader internet. Menno even takes us back to the rise of home computers in the 1980s, drawing some surprising parallels with today’s AI revolution. Of course, the episode doesn’t shy away from the tricky topics either, like privacy concerns and the risk of the internet becoming a bit too uniform.

We bespreken de paradoxen van AI, de dopaminefabriek van het experimenteren met nieuwe tools, en de impact van AI op design, muziek en het internet. Menno neemt ons mee terug naar de opkomst van homecomputers in de jaren ’80, en trekt verrassende parallellen met de AI-revolutie van nu. Natuurlijk komen ook de minder leuke kanten aan bod: van privacyvraagstukken tot de dreiging van eenheidsworst op het internet.

Verder hoor je:

• Waarom Zwolle volgens Menno de leukste stad van Nederland is

• Hoe AI het creatieve proces versnelt, maar niet altijd vervangt

• De toekomst van advertising, contentcreatie en klantreizen met AI

• Retro gaming, bedroom coders en de kracht van menselijke creativiteit

Een aflevering vol slimme inzichten, nostalgie en scherpe discussies – met een gast die niet op zijn mondje is gevallen. Luister mee en ontdek waarom AI eigenlijk best wel leuk is!

The Current AI Landscape

AI chatbots and assistants have experienced insane growth since late 2022, with some platforms gaining over 100 million users in mere months. As of May 2025, the competition between major AI players has intensified, creating a rich ecosystem of options for users with varying requirements and preferences. These powerful tools can generate human-like text, process images, create code, analyze data, and increasingly perform complex reasoning tasks that were once exclusive to human intellect.

Understanding AI Foundations

Before diving into specific platforms, it's essential to understand that modern AI assistants operate on large language models (LLMs) trained on vast datasets of text and other media. Different models have varying capabilities in multimodality (handling text, images, audio, and video), context length (how much information they can process at once), and specialized skills like coding or data analysis. For the average user, most of these things won’t make a massive difference and it all comes down to personal preference. But, it is important to at least know the capabilities of others.

Reasoning is another term that gets used a lot within AI. The easiest way to understand reasoning is just by its name: “Reasoning”. Instead of giving you a reply solely based on your input, it actually reasons and analyzes the data before responding. This will give you more thought through responses. This is mostly used for more complex questions that involve different steps and data.

Major AI Platforms Compared

ChatGPT by OpenAI

OpenAI's ChatGPT is one of the most popular models currently. It’s easy to use interface with great capabilities make for a perfect match for beginners. The current GPT-4o model is your standard pre-trained LLM and will be used for most cases. The o3 model is the reasoning model from ChatGPT, which can be used for the more intricate questions. The other models function somewhat similar, the best thing to note is that any model with GPT in front is (Generative Pre-Trained Transformer) pre-trained and any other model likely includes reasoning.

Link: chatgpt.com

Google's Gemini

Google has been making big strides to catch up to some of the leading AI company’s and they have done so quite well. It works the exact same as ChatGPT where the 4o model is similar to Gemini 2.0 Flash and the 2.5 models are close to o3. Gemini does offer great integration with different Google services.

A small mention to Gemini 1.5 Flash that can process 3000 images at a time, analyze an hour long video and listen to hours of audio. This is however only accesible through the ai studio from Google.

Links:

• gemini.google.com

• aistudio.google.com (for Gemini 1.5 Flash)

Claude by Anthropic

Anthropic's Claude is another strong contender, especially if you care about AI safety and getting more thoughtful responses. Their latest, Claude 3.7 Sonnet, is pretty capable. It’s good at breaking down complicated questions and can show you its reasoning, which is handy. If you're looking for an AI that's a bit more transparent in how it "thinks" or handles nuanced tasks, Claude is worth a look. They also have options for developers.

Link: claude.ai

Le Chat by Mistral AI

Mistral AI, a company from France, offers Le Chat. It's known for being fast, especially with its "Flash Answers" feature. Beyond just chatting, Le Chat is pretty good at handling documents – think PDFs, spreadsheets, and even pulling text from images. If you need quick processing or document analysis, this could be your go-to. They also offer tools for image creation and coding.

Link: chat.mistral.ai

Perplexity AI

Perplexity AI is a bit different; think of it more like a super-powered research assistant. When you ask it questions, it doesn't just give you an answer, it also shows you the sources where it found the information. This is awesome if you're doing homework, writing a paper, or just want to check where the AI is getting its facts. If you need answers with backup, Perplexity is great.

Link: perplexity.ai

Additional AI Models

Deepseek
Developed in China, Deepseek made headlines in early 2025 with rapid advancements in both reasoning and multimodal capabilities. Its v3.1 model particularly stands out for coding tasks and STEM problem-solving.

Link: chat.deepseek.com

Qwen
This emerging model offers unique strengths in multilingual processing (handling 10+ languages) and specialized domains like audio analysis and financial forecasting. The Qwen2.5-Max version even understands tables and spreadsheets for data-heavy workflows.

Link: chat.qwen.ai

Grok by xAI
xAI's Grok is making waves with its no-nonsense approach and real-time knowledge from the X platform. The latest Grok 3 model shines in tackling complex math and science questions, even offering a "Think" mode that shows its step-by-step reasoning. While currently limited to X Premium+ subscribers, it's positioning itself as a strong alternative to ChatGPT for research-heavy tasks.

Link: x.ai

Llama by Meta
Meta's Llama stands out as the most powerful open-source option available. The massive Llama 3.1 405B model rivals paid alternatives in coding and multilingual tasks, while smaller versions like the 70B offer solid performance for everyday use. Developers appreciate its flexibility – you can fine-tune it for specific needs or run it locally for privacy-focused projects.

Link: llama.com

Running Your Own AI Model

For those interested in greater customization, privacy, or specialized applications, running your own AI model locally or on private infrastructure represents an increasingly accessible option. Our video explores the practical steps, hardware requirements, and software frameworks needed to deploy personal AI systems.

Tutorial Link:

Conclusion

At the end of the day, it's all down to personal preference. What works for me might not work for you, and that's totally fine.

The cool thing about AI in 2025 is that there's something for everyone, whether you're trying to write a book, debug some code, dig deep into research, or just mess around and see what these things can do. This video's meant to give you the starting points, but the real magic happens when you start playing around with them yourself.

Each of these platforms has its own quirks and strengths, and honestly, the best way to figure out which one clicks for you is to jump in and try them out. Pay attention to which interface you like better, which pricing model makes sense, and what features you find yourself using the most. AI is always changing, so stay curious, keep experimenting, and you'll find the perfect tool (or tools) for your needs.

Just remember: It's not about finding the "best" one, it's about finding the one that helps you do your thing.

In the rapidly advancing realm of artificial intelligence, video generation technology has reached astonishing new heights. Let's explore three revolutionary platforms that are redefining what's possible in motion picture creation, making professional-quality video production accessible to creators of all skill levels.

Higgsfield AI: Cinema-Quality Motion Control

Founded by former Snap AI head Alex Mashrabov, Higgsfield AI has emerged as a standout player in the AI video generation space. They are most dominant on Image to Video generation.

Key Features:

• Suite of 56 dynamic camera styles including bullet time effects, crash zooms, and dolly shots

• Single-image reference system that transforms still images into fluid cinematic sequences

• Physics-based motion that adds realistic weight and momentum to movements

• Intuitive interface that requires no technical expertise to create professional-looking shots

Higgsfield AI offers tiered pricing with a free plan for experimentation, a Standard plan at $14.90 per month, and more comprehensive Pro options for serious creators. The platform outputs 3-5 second clips at 720p resolution, making it ideal for social media content and scene-specific productions.

Runway Gen-4: Crazy Consistency

Runway's Gen-4 model represents a significant leap forward in maintaining visual coherence throughout AI-generated videos, addressing one of the industry's most persistent challenges.

Standout Features:

• Character consistency that maintains the same person across different scenes and angles

• Multi-perspective capabilities allowing generation of the same moment from different camera positions

• Lighting and mood preservation that maintains the visual tone throughout the sequence

• Two speed options: standard Gen-4 for maximum quality and Turbo for faster iteration

The platform supports multiple aspect ratios (16:9, 9:16, 1:1, 4:3, 3:4, 21:9) with output at 24fps-true cinema framerate.

Sora's Video Model: Length and Realism Redefined

OpenAI's Sora represents perhaps the most ambitious video generation system to date, with capabilities that were considered science fiction just months ago.

Highlights:

• Extended duration capabilities of up to 20 seconds at 1080p resolution

• Remarkably accurate physics simulation that captures real-world movement patterns

• Ability to extend, remix, and blend existing video content seamlessly

• Advanced storyboard tool for precise frame-by-frame creative control

After an initial research preview earlier this year, Sora is now fully available to ChatGPT Plus, Team, and Pro users through a dedicated interface at sora.com. The newer "Sora Turbo" model offers significantly faster generation times without compromising quality.

These groundbreaking AI video tools are very versatile. From social media marketers needing eye-catching short-form content to independent filmmakers looking to visualize concepts before shooting, these tools offer compelling solutions that would have required entire production teams just a few years ago.

As these technologies continue to evolve, we can expect even more sophisticated capabilities, longer durations, and higher resolutions-potentially reshaping entire industries from advertising and entertainment to education and corporate communications.

Thanks for reading! For more tech news:

Follow us on Instagram: @mach8.io and @mach8.labs
Follow us on TikTok: @mach8.io

In deze aflevering van Techdives schuiven Bas, Eelco en Marc aan voor een open gesprek over AI agents. Geen sciencefiction, geen revolutie – deze digitale assistenten draaien al een tijdje mee en maken het werk op de achtergrond gewoon nét wat makkelijker.

Met een demo van United Playgrounds laten we zien hoe verschillende bureaus hun eigen AI agent inzetten, van het stroomlijnen van sollicitaties tot het ondersteunen van creatieve processen. We bespreken waar je ze tegenkomt, hoe ze je werkdag soepeler maken en waarom je er eigenlijk niet meer omheen kunt – zonder dat het allemaal rocket science is.

Kortom: een aflevering over technologie die inmiddels gewoon bij het meubilair hoort, maar toch steeds weer verrast.

No fluff, just the features you need to know. Whether you're a designer, marketer, educator, or just Canva-curious, this update is opening new creative horizons.

👉 Watch now and see why Canva is not just keeping up—it’s leading the future of design.

Step 1: Crafting the Script with ChatGPT

First, you’ll need a compelling script for your Rituals ad campaign. ChatGPT is perfect for brainstorming and writing your script, but remember to always double check the output. The script is arguably the most important part of a successful video.

Here’s how you can use ChatGPT:

1. Provide a prompt: Start by describing the Rituals brand and the message you want to convey. For instance, you could ask ChatGPT to generate a script focusing on the luxury and self-care aspects of Rituals’ products.

   Example: "Write a 60-second ad script for Rituals that highlights the luxury of their skincare products and the importance of self-care in everyday life."

2. Refine the script: Once you have the initial script, you can ask ChatGPT to tweak the tone or length, ensuring it fits the brand’s image.

3. Finalize the script: Once satisfied, you have your script ready for the next stage!

Step 2: Generating Visuals with MidJourney, Sora, or Gemini

Now that you have the script, it’s time to think about visuals. Tools like MidJourney, Sora(ChatGPT 4o), and Gemini are fantastic for creating high-quality images for your ad. We are going to be using these tools to create a First and a Last frame of a scene. We can later use these to generate our video’s.

• Map out your scenes from the script and think of clever ways to write a prompt to match this shot.

• Generating Time! Generate images to get a feel of the different shots and potentially even change the prompt. Make sure to create different iterations to get a bigger picture of options.

• Decide which frames you are going to be using for the add.

Once you have the images, you can refine and enhance them in Photoshop, readying them for animation. You might want to choose a slightly different colour or remove a object. This can all be done in Photoshop

Step 3: Separating Side Assets for More Control.

Once you have your scenes mapped out and the shots generated, it is important to also take a look at any subjects within the video. In the example of Rituals, the fish had to have more control.

1. Separate Assets: Use any of the previously mentioned image generators to generate imagery of any items/subjects.

2. Meshy: Meshy is used to create any 3D models of objects that you might need in the video. You can then position these in every shape or form.

3. Make Them Transparent By generating them with a green screen background, they can be easily separated and made transport to make the implementation process easier.

4. Finalization: While you are still in this stage, make sure to match the lighting and shading of the generated shots. This will make everything look more cohesive in the end. (This is also the time to finalize position and rotation).

Step 4: Generating the Footage with Kling and Runway

Now comes the fun part—bringing your visuals to life using AI-powered animation tools.

• Kling/Runway: Use Kling or Runway to create smooth transition between your first and last frame.

• Continuation: If you want to create longer shots, you can use the ending frame of the first scene as the starting scene of the next.

Make sure the footage flows with the rhythm of your script, pacing out key moments where you want emphasis on certain visuals.

Step 5: Final Edits in Premiere Pro or CapCut

Once your animation footage is ready, it’s time to stitch everything together in Premiere Pro or CapCut.

1. Import your footage: Bring all your animated clips into your editing software.

2. Arrange your scenes: Follow the structure of your script and start placing your visuals in sequence.

3. Add voiceover and music: Overlay the voiceover based on the script you created earlier. Add soft, ambient music that fits the luxurious and relaxing vibe of Rituals.

4. Adjust timing and transitions: Use smooth transitions to connect the scenes. Ensure the pacing feels right and that the visuals complement the voiceover.

5. Export your video: Once you’re happy with the final result, export the video in the necessary format for sharing and distribution.

The Final Result

By following these steps, you’ll have created a beautiful Rituals ad campaign video using cutting-edge AI tools and animation techniques. The result will be a high-quality video that reflects the luxury and self-care ethos of the Rituals brand—engaging, soothing, and visually captivating.

And there you have it—an end-to-end workflow using the best tools available today to create an ad campaign video from scratch. Don't forget to show off your work to the world!

Thanks for reading! for more tech news:

Follow us on Instagram: @mach8.io and @mach8.labs

Follow us on TikTok: @mach8.io

In the ever-evolving world of artificial intelligence, image generation capabilities are reaching new heights. Let's explore the latest advancements from three industry leaders that are reshaping the landscape of visual creativity.

ChatGPT's GPT-4o: A New Era of Image Manipulation

On March 26, 2025, OpenAI unveiled a major upgrade to ChatGPT's image generation capabilities, powered by the GPT-4o model5. This groundbreaking update allows ChatGPT to not only create images from scratch but also modify existing ones with unprecedented accuracy and detail.

Key Features:

• Native image creation and editing within ChatGPT

• Ability to transform and 'inpaint' details in images, including those with people

• Longer processing time for more accurate and detailed results

GPT-4o's image capabilities are currently available to ChatGPT Pro subscribers, with plans to roll out to Plus and free users soon. This advancement positions ChatGPT as a versatile tool for both text and visual content creation, offering users a seamless experience for their creative projects.

Examples:

Reve Image 1.0: Stunning Visuals at Your Fingertips

California-based AI startup Reve AI Inc. has introduced Reve Image 1.0, a text-to-image generation model that's taking the internet by storm2. This free preview model excels in prompt adherence, typography, and aesthetics, making it accessible to users without advanced prompt engineering knowledge.

Highlights:

• Available for free at preview.reve.art

• User-friendly interface for generating images from text prompts

• Exceptional performance in realistic image creation

Reve Image 1.0 has quickly gained popularity among AI enthusiasts, with users sharing their creations on social media platforms. Its ability to generate highly detailed and imaginative images has positioned it as a top contender in the text-to-image leaderboard.

Examples:

Google's Gemini 2.0 Flash: Multimodal Mastery

Google has upped the ante with Gemini 2.0 Flash, a groundbreaking update to their AI model that's set to transform image creation and editing3. This multimodal powerhouse combines text understanding with advanced image manipulation capabilities.

Standout Features:

• Edit specific parts of an image without altering the rest

• Maintain character consistency across different scenes

• Generate step-by-step visual guides

• Use conversational prompts for detailed adjustments

Gemini 2.0 Flash is available for free through Google's AI Studio, allowing users to experiment with its powerful image generation and editing tools. This update solidifies Google's position as a leader in multimodal AI technology.

These advancements in AI-powered image generation are opening up new possibilities for creators, designers, and businesses alike. Whether you're looking to enhance your visual content, streamline your design process, or explore new realms of creativity, these tools offer unprecedented capabilities at your fingertips. As AI continues to evolve, we can expect even more exciting developments in the world of visual creation and manipulation.

Examples:

Thanks for reading! For more tech news:

Follow us on Instagram: @mach8.io and @mach8.labs

Follow us on TikTok: @mach8.io

• Ollama

• Ollama Github (Raw Information)

• Ollama Modelfile

• Visual Studio Code

We analyseren hoe bedrijven als Meta en Anthropic hun koers verleggen met onderzoekers aan het hoofd en waarom dit een fundamentele verschuiving betekent in hoe technologie wordt ontwikkeld. Is dit de sleutel tot grensverleggende innovaties, of zet het de deur open voor onbeheersbare risico’s? En hoe reageren investeerders, klanten en concurrenten op deze verschuiving?

Daarnaast bespreken we wat dit betekent voor de toekomst van leiderschap in tech. Zullen we steeds vaker zien dat de knapste koppen direct de touwtjes in handen krijgen, zonder tussenkomst van traditionele managers? Of komt er een moment waarop de balans tussen wetenschap en commercie zoekraakt?

Bereid je voor op een diepgaande en prikkelende discussie over de toekomst van tech, macht en de rol van visionairs in de digitale revolutie!