PLUS: Meta’s ‘Segment Anything’ for audio and Linus Torvalds blesses AI
Good morning
OpenAI has just rolled out a major update to ChatGPT Images, introducing a new flagship model that makes creating and editing images more precise and dramatically faster.
The upgrade effectively turns the platform into a powerful creative studio with a new level of control. But does this shift toward advanced in-platform editing set a new bar for what we should expect from all AI image tools?
In today’s Next in AI:
ChatGPT's new image editing features
Meta's 'Segment Anything' audio model
Linus Torvalds' backing of AI for code
The Junior 'Jobpocalypse'
ChatGPT's Image Upgrade
Next in AI: OpenAI released a new version of ChatGPT Images powered by a flagship model that makes creating and editing images faster and more precise than ever.
Decoded:
The new model excels at targeted edits and preserves original details, allowing you to change specific elements in a photo while keeping the lighting, composition, and overall style consistent.
Performance gets a major boost, with image generation now up to 4x faster and the new model, GPT Image 1.5, available to developers through the API.
Beyond simple edits, the new Images experience in ChatGPT enables creative transformations, letting you turn photos into movie posters or apply complex stylistic filters with simple text prompts.
Why It Matters: This update moves ChatGPT beyond basic image generation, turning it into a powerful and intuitive creative studio for professionals and hobbyists alike. This new level of control and speed sets a higher bar for what users can expect from AI image editing tools.
Segment Anything... for Audio
Next in AI: Meta just applied its "segment anything" magic to sound, releasing SAM Audio, a new model that isolates specific sounds from complex recordings using simple text, visual, or time-based prompts.
Decoded:
It works with multimodal prompts, allowing you to isolate audio by typing a description like "singing voice," clicking on a guitar in a video, or highlighting a time segment to remove a recurring noise.
The system is powered by the new Perception Encoder Audiovisual model, which aligns video frames with audio to precisely understand what is being seen and heard at the same time.
To support the launch, Meta also released new evaluation tools, including a comprehensive benchmark and an AI-powered judge to measure the quality of separated audio against human perception.
Why It Matters: This technology makes high-level audio editing accessible to creators without requiring professional sound engineering skills. It also paves the way for smarter creative tools and significant improvements in accessibility features like advanced noise cancellation.
Linux Czar Blesses AI
Next in AI: Linux creator Linus Torvalds, a notable skeptic of tech hype, has become a "huge believer" in using AI as a tool for maintaining code. The topic is now so important that new AI tooling was a key discussion at the recent Kernel Maintainer Summit, signaling a major shift for the project.
Decoded:
Torvalds is far more interested in AI for automated patch checking and code review, viewing it as a powerful new tool rather than a way to generate code.
His confidence grew after an internal experiment where an AI reviewing a code merge not only found his objections but also added new objections of its own.
He places AI's impact as an evolution, noting that while compilers were a 1,000x boost, AI might add another 10x or 100x on top of that.
Why It Matters: An endorsement from Torvalds gives immense credibility to using AI assistants in mission-critical, open-source software development. This practical application could accelerate the adoption of similar code-auditing tools across the entire tech ecosystem.
The Junior 'Jobpocalypse'
Next in AI: A wave of AI-driven automation is gutting entry-level tech jobs, as tasks like debugging and routine software maintenance are increasingly handled by algorithms instead of recent graduates.
Decoded:
Hiring for new graduates at big tech companies has declined by more than 50% over the last three years, according to a SignalFire talent report.
This isn't just a Silicon Valley trend; Indian IT firms have reduced entry-level roles by up to 25% as they adopt AI and automation.
Employer expectations are shifting, with companies now seeking graduates who can also handle project management, sales, and customer-facing responsibilities right out of college.
Why It Matters: The era of relying on a computer science degree alone to secure a foothold in the tech industry is rapidly closing. Future professionals must now build a T-shaped profile, blending deep technical expertise with strong communication and business skills to stay relevant.
AI Pulse
Urban VPN was exposed for harvesting and selling user conversations from major AI platforms like ChatGPT and Claude through its "privacy" browser extensions, affecting over 8 million users.
Google released A2UI, a new open-source protocol that allows AI agents to securely generate rich, interactive user interfaces across multiple platforms without executing arbitrary code.
AIsbom launched a specialized security scanner for ML models that performs deep binary introspection to detect malware and license violations hidden inside serialized weights.
SignalFire reported that entry-level hiring at big tech companies has dropped by over 50% in the last three years as AI automates tasks previously handled by junior software engineers.