Author HubMatthias Bastian

Matthias is the co-founder and publisher of THE DECODER, exploring how AI is fundamentally changing the relationship between humans and computers.
AI research

Wait a minute! Researchers say AI's "chains of thought" are not signs of human-like reasoning

AI in practice
Update

US Copyright Office says fair use does not cover AI trained on "vast troves of copyrighted works

AI research

OpenAI's o3 is less AGI than originally measured

Ad
AI in practice
Short

Claude, the chatbot from AI startup Anthropic, will soon be able to handle ten times more content in its "Projects" feature. If a user exceeds the regular limit, a new "Retrieval" mode will automatically turn on to expand the chatbot’s memory. This upgrade will roll out to all paid Claude plans in the coming days.

AI in practice
Short

Google has rolled out another update to its flagship AI model, Gemini 2.5 Pro. The latest version brings modest improvements across a range of benchmarks and maintains top positions on tests like LMArena and WebDevArena. According to Google, the new release also handles challenging programming, knowledge, and logic tasks better than before. The model also "can be more creative with better-formatted responses."

Deepseek's R-1 still offers the best price-performance ratio, but Gemini 2.5 Pro clearly leads among proprietary Western AI models, staying ahead of OpenAI's offerings. | Image: Google

Developers can access the updated Gemini 2.5 Pro in preview through Google AI Studio and Vertex AI. Google plans to promote this preview to a stable, generally available version in the coming weeks. The update is also rolling out to the Gemini app starting today.

Ad
Ad
AI in practice
Short

Anthropic has added three premium features to Claude Pro, previously only for Max, Team, and Enterprise users. Claude Code allows command-line coding tasks. A new integration tool connects Claude to apps like Jira, Zapier, and Asana. The research feature now offers deep analysis from web and data sources, similar to deep research capabilities from Google and OpenAI. All features are included in the current Pro plan at no extra cost.

Claude's version of "Deep Research" is now part of the Pro plan. | Image: Screenshot THE DECODER
AI and society
Short

Amazon MGM Studios is reportedly producing a film about OpenAI's 2023 leadership crisis. Titled "Artificial," the movie will be directed by Luca Guadagnino and written by Simon Rich, who is also producing. The story centers on Sam Altman's temporary removal and swift return as OpenAI CEO. Andrew Garfield is in early talks to play Altman, with Monica Barbaro as CTO Mira Murati and Yura Borisov as co-founder Ilya Sutskever. Filming is planned for summer 2025 in San Francisco and Italy, according to Deadline and The Hollywood Reporter.

Ad
Ad
AI and society
Short

In 129 documented cases across 12 countries, lawyers have submitted fake legal content generated by AI tools like ChatGPT into court proceedings, according to a database by international law expert Damien Charlotin. Only cases with official disciplinary action or court rulings are included - this is definitely just the tip of the iceberg. The database is updated regularly and aims to highlight the dangers of using AI tools without proper checks in legal contexts. Charlotin, who teaches AI and law at HEC Paris, added over 30 new cases in May 2025 alone.

Ad
Ad
AI and society
Short

TikTokers are reversing the usual AI confusion by pretending their real videos are made by Google's new video model, Veo 3. Indie band Kongos posted a clip of a live performance labeled as AI-generated. Rapper GameBoi Pat did the same with an older song, while others like Darden Bela made similar claims with existing videos. Some users act like "prompts", or joke that AI made them gamble or drink. "Real or Veo" quizzes are also trending. It highlights how blurred the line is between real and AI videos. Google reports that millions of AI videos have been generated using Veo 3 in just a few days.

Google News

Bank details

IBAN: DE87 1203 0000 1086 0070 75
Account holder: DEEP CONTENT GbR
Purpose: Support THE DECODER
AI research

New Othello experiment supports the world model hypothesis for large language models

AI in practice

ChatGPT might be draining your brain, MIT warns - what ‘cognitive debt’ means for you

AI research

Meta's latest model highlights the challenge AI faces in long-term planning and causal reasoning

Google News