Today, OpenAI is ushering in a new era of AI with a real game changer: GPT-5 will soon launch with unlimited, free access, combining several AI features into a single, smart model. While the last "legacy" chatbot, GPT-4.5 (Orion), gets its finishing touches, this consolidation promises a much simpler world for users and developers.

Adobe is also getting in on the act - the San Jose, California-based company has unveiled the beta version of its new AI video generator. In a word, it is impressive.

Enjoy reading and diving into the future of AI!

Marcus Schuler

OpenAI: GPT-5 comes with unlimited free access

Major strategic U-turn at OpenAI in San Francisco. CEO Sam Altman is effectively burying GPT-o3 in favor of a new model called GPT-5, which will combine different technologies into a single product.

The major update to its AI offering marks the launch of GPT-5 with unlimited free access for users on the free tier. This departs from the previous model, where separate tools - such as Voice, Canvas, Search, and Deep Research - will soon be integrated into a single AI system. With this integration, GPT-4.5 (Orion) will be the last standalone model to be released shortly.

In a blog post, Altman said: "We want to communicate our planned roadmap better and significantly simplify our product offering. We want to make AI easier to use; we know how complicated our model and product offering has become. We hate model selection in ChatGPT as much as our users do and want to return to the magic of unified intelligence".

One thing is clear: this consolidation will significantly simplify the AI landscape. For the developer community, it eliminates the need to integrate different models. Individual users will benefit from direct access to advanced AI capabilities without worrying about the technical overhead. In addition, integration and deployment will be easier for developers as the unified system reduces the fragmentation that has hindered the ecosystem.

Overall, this strategic move by OpenAI aims to improve the user experience, promote cross-platform efficiency, and support broader adoption by removing redundant barriers to deploying AI tools.

Why it matters:

• Optimized innovation: Combining different AI capabilities in a single system enables more agile product development. OpenAI can iterate faster and incorporate innovative features by reducing the need to manage multiple models.
• Improved user engagement: Unrestricted access to GPT-5 for free tier users lowers the barrier to entry and can encourage widespread adoption. This approach lays the foundation for a larger, more engaged community whose feedback can drive future product improvements.
• Optimized ecosystem for developers: A single, consistent platform simplifies integration and deployment for developers. By minimizing complexity, developers can build on top of the system more easily and quickly.

There is also a significant new development in deep research. OpenAI wants to make the tool, developed for the scientific community, available to a broader audience. The service will be available to ChatGPT Plus users ($20 per month) with 10 users per month and to free users with two users per month.

What's new in GPT-4.5?

But before GPT-5 sees the light of day, the next version codenamed “Orion,” GPT-4.5, is out. Here are the main changes:

• Enhanced specifications document: The new specs have expanded from approximately 10 to 63 pages. They offer detailed and comprehensive guidelines on how the AI should operate across various areas, including controversial topics and user customization.
• Incorporating current AI ethics: The updated guidelines address issues such as gender identity.
• Sex and Pornography: In the future, there will be an "adult mode" for sex and eroticism. Harmful materials like revenge porn and deepfakes will still be banned.
• Factual tone: The new model is designed to provide consistent and factual responses moving forward, encouraging more constructive criticism instead of excessive praise. The goal is to have fewer friends and more thoughtful colleagues.
• Establishing a hierarchical chain of command and open sharing: The specification establishes a chain of command in which platform-level rules take precedence over developer guidelines, followed by user preferences. It is released under a CC0 license that permits free modification and adoption by the broader AI community.

Read on, my dear:

• OpenAI: OpenAI Model Spec
• OpenAI: Sharing The Latest Model Spec
• TechCrunch: OpenAI cancels its o3 AI model in favor of a ‘unified’ next-gen release
• engadget: OpenAI will offer free ChatGPT users unlimited access to GPT-5
• The Verge: OpenAI is rethinking how AI models handle controversial topics
• VentureBeat: OpenAI CEO Sam Altman shares plans to bring o3 Deep Research agent to free and ChatGPT Plus users
• AXIOS: OpenAI's new roadmap: big releases in "weeks" and "months"

AI Photo Of The Day:

Prompt:
beautiful woman photograph, shot with Pentax film camera, huge chromatic aberration, prisms flares, much grain and noise, beautiful eyes, sharp sight, finely detailed sAIn, eyebrows commercial, predator eyes, passion and self confidence.

Adobe releases beta of new AI video generator

Users can create five-second video clips at 1080p resolution. The tool provides text-to-video and image-to-video features that enhance the generated content. Adobe has also launched new subscription plans for its Firefly models, offering unlimited access to image creation capabilities. The new video generator is now available in public beta.

Why it matters:

• The new video generator is integrated into the redesigned Firefly web application and connects directly to Creative Cloud applications, including Premiere Pro, Photoshop, and Express. This facilitates a seamless transition from AI-generated content to traditional editing environments, ensuring production-ready material safe for commercial use due to licensed training data.
• With a range of text-to-video and image-to-video capabilities, users can now create video content using either descriptive text or a reference image while maintaining detailed control over creative parameters such as style, camera angle, and movement. This dual approach signifies a significant advancement in creative expression and rapid ideation.
• Enhanced quality and future scalability: The generator now outputs video at 1080p/24fps, representing a significant improvement over the original 720p quality. There are plans to introduce faster, lower-resolution ideation models, along with a future 4K option. Adobe is positioning itself for the rapidly growing market for high-resolution AI video.

Read on, my dear:

• Adobe: Meet Firefly Video Model: AI-Powered creation with unparalleled creative control
• Adobe: Adobe Expands Generative AI Offerings Delivering New Firefly App with Industry’s First Commercially Safe Video Model
• reuters: Adobe launches AI video tool to compete with OpenAI
• Thurrott: Adobe Brings Video AI Model to Firefly, Premiere Pro

Top 10 AI tools for creating cool presentations fast

• Gamma - Create stunning presentations, documents, and websites without any design or programming skills. Gamma
• SlidesAI - AI-powered presentations in seconds. SlidesAI
• Tome - The leading AI-based sales and marketing research and presentation platform. Tome
• AutoSlide - From idea to finished presentation in seconds. AutoSlide
• MagicSlides - PowerPoint presentations can be made by simply entering a topic, YouTube link, PDF, or text. MagicSlides
• DesignerBot - It helps you brainstorm with text and images. DesignerBot
• MotionIt AI - AI-generated elegant slides from keywords or documents - including automatic formatting and image selection. MotionIt AI
• SendSteps AI - It takes care of copywriting, design, and storytelling so you can focus on presenting. Sendsteps AI
• Storytelling - Business presentations with compelling storytelling - powered by AI. Storyd
• PitchBob - A tool for creating pitch decks, business plans, and landing pages specifically for startups. PitchBob

AI and tech news

Open source or obsolete: the West must innovate to beat China in AI

Former Google CEO Eric Schmidt has warned that Western nations need to adopt open-source AI models to avoid falling behind China, following the launch of DeepSeek’s efficient R1 language model. He criticized the largely closed approach of leading AI systems in the U.S. and cautioned that this could hinder scientific research at Western universities. Schmidt announced a $10 million investment in an AI security research program to address these challenges and called for closer cooperation with China on AI security issues.

EU scraps AI liability directive

The European Commission has removed the AI liability directive from its 2025 work program after facing significant criticism of the regulation at the AI Action Summit in Paris. The decision comes amid vocal disapproval from figures such as US Vice President JD Vance and the realization that the directive has become redundant after adopting the EU's AI law.

Snowflake expands AI tools by partnering with Anthropic

Snowflake has teamed up with Anthropic to integrate AI agents into enterprises for enhanced data analysis while maintaining security. The new Cortex Agents platform will assist organizations in securely analyzing both structured and unstructured data. The aim of this collaboration is to establish AI as a crucial tool for boosting productivity across various teams within an organization.

Google Gemini tampering: Hackers implant persistent false memories via delayed injection

A new hacking technique enables attackers to corrupt the long-term memory of Google's Gemini chatbot through delayed injection. Researchers have demonstrated that by manipulating user interactions with untrusted documents, they can make Gemini permanently retain false information. Google has acknowledged the issue but claims that the overall risk is low, while the researchers caution that corrupting AI memory could result in misinformation.

Ultra-high neutrino record: new insights into cosmic extremes

In the depths of the Mediterranean Sea, physicists have discovered the most energetic neutrino ever measured on Earth. The KM3NeT underwater telescope has indirectly detected a neutrino with an energy of approximately 220 million trillion electron volts, far surpassing all previous records. This groundbreaking discovery opens a new avenue for research into cosmic accelerators and high-energy processes in the universe.

How to…

…Use ChatGPT in 2025

Robert Brown

Novato

Tech journalist. Lives in Marin County, north of San Francisco. Got his start writing for his high school newspaper. When not covering tech trends, he's swimming laps, gaming on PS4, or vibe coding through the night.