ChatGPT can now read some of your Mac’s desktop apps

OpenAI has introduced a new feature for its ChatGPT desktop app on macOS, enabling the chatbot to work directly with coding apps like VS Code, Xcode, TextEdit, Terminal, and iTerm2. This feature, called "Work with Apps," allows developers to send sections of their code to ChatGPT automatically as context for their prompts, eliminating the need to copy and paste manually. While the chatbot can suggest code changes or additions based on the context provided, users still need to paste the generated code back into their coding environment.

The feature relies on macOS’s accessibility API, which reads text and translates it for ChatGPT. For some apps, such as VS Code, additional extensions are needed. However, the feature is limited to text-based interactions and cannot interpret visual elements like images or videos. By default, ChatGPT can process the last 200 lines of code or the content of the active window as input, and users can highlight specific sections to focus the AI’s attention.

OpenAI emphasized that this is not a fully autonomous AI agent but a stepping stone toward creating more collaborative and agentic systems. The company plans to expand the feature to include other text-focused tools in the future. The rollout is currently limited to macOS Plus and Teams users, with Enterprise and Education versions expected soon. Windows support has not yet been announced, though OpenAI hinted at broader integration plans, especially as competition in the AI agent space intensifies with rivals like Anthropic and Google exploring similar functionalities.

It’s time to change the way we build digital products

Airtable ProductCentral is a complete operating system for Product teams, built on Airtable's powerful platform. It manifests the best practices of world-leading Product teams that Airtable has partnered with for years.

Airtable ProductCentral unifies your entire product organization. Synthesize customer feedback, visualize investments across all product teams, allocate resources based on real-time data, and translate every initiative into financial impact.

New AI Tool Tracks Your Steps by Reading the Bacteria You Carry

A research team from Lund University in Sweden has developed an innovative AI tool called Microbiome Geographic Population Structure (mGPS) that can trace recent human locations by analyzing microorganisms. This breakthrough leverages the unique "geographic fingerprints" of bacteria found in various environments, such as urban centers, beaches, and forests. The tool has potential applications in medicine, epidemiology, and forensics by identifying microbial signatures linked to specific geographic origins.

Unlike human DNA, the microbiome changes as individuals interact with different environments. By analyzing these changes, mGPS can trace the spread of diseases, identify infection sources, and even support forensic investigations. The AI model was trained using vast datasets, including over 4,100 microbiome samples from public transport systems in 53 cities, 237 soil samples from 18 countries, and 131 marine samples from nine bodies of water. The tool successfully pinpointed the city of origin for 92% of urban samples, with remarkable precision even in specific locations like subway stations and kiosks.

This groundbreaking research underscores the potential of microbial analysis for advancing forensic science. According to lead researcher Eran Elhaik, future efforts aim to map entire city microbiomes, unlocking a deeper understanding of how microorganisms interact with environments and providing new insights into urban ecosystems, public health, and crime-solving tools.

Not even Spotify is safe from AI slop

Fake AI-generated albums have been appearing on Spotify, targeting real artists and undermining their music. Bands like HEALTH, Annie, and Standards have found unauthorized releases on their official artist pages, often with AI-generated music and artwork. These fake albums are uploaded through distributors who handle metadata and licensing, allowing fraudsters to sneak their content onto legitimate profiles. While Spotify removes some of this content eventually, artists often face delays and struggle to get support. These fake releases not only confuse fans but also divert attention and revenue from authentic music.

The root of the issue lies in the honor-based system used by distributors and streaming platforms. Fraudsters exploit this by uploading AI-generated content and collecting royalties meant for the real artists. For example, a suspicious “Gupta Music” label uploaded over 700 fake albums with one-word artist names, like “Bedroom” and “Culture,” all hosted under legitimate profiles. Although Spotify removed this content after cutting ties with the distributor Ameritz Music, it highlights how vulnerable the system is to abuse. Similarly, scams like bot-driven streams and typo-based artist impersonations (e.g., “Arriana Gramde”) show how far fraudsters will go to exploit the platform.

This type of fraud is lucrative because small payouts from many fake streams can add up to significant earnings, often siphoned off through fake labels and compromised accounts. While Spotify claims to invest in fraud prevention, gaps in metadata verification and enforcement persist. Artists, like those targeted by “Gupta Music,” often rely on distributors and Spotify to take action, but the process is slow, leaving the fake content online for weeks or months.

The rise of AI has made this type of fraud easier and more widespread, accelerating a problem that already existed. Fraudsters previously altered or digitized old music to exploit royalties, but now AI tools streamline the creation of fake tracks. For platforms like Spotify, this influx of AI-generated content could degrade user trust and the overall experience, just as spam and misinformation have affected other tech platforms. Without better moderation and stronger safeguards, both artists and listeners will continue to be victims of this broken system.

Other stuff

• OpenAI Email Archives (from Musk v. Altman)

• Trump’s Anti-Regulation Pitch Is Exactly What the AI Industry Wants to Hear

• Gemini AI tells the user to die

• Diagrams AI Can, and Cannot, Generate

• ‘It gets more and more confused’: can AI replace translators?

• AI could cause ‘social ruptures’ between people who disagree on its sentience

• AI isn’t about unleashing our imaginations, it’s about outsourcing them. The real purpose is profit

• How AI Could Break the Career Ladder

• Daisy, The ‘AI Granny’ Outwitting Scammers

Hope you enjoyed today's newsletter

⚡️ Join over 200,000 people using the Superpower ChatGPT extension on Chrome and Firefox.