GPT-5.4 Scores 95% on USAMO 2026: AI Mathematical Reasoning Hits a New Ceiling
OpenAI's flagship model essentially saturates the US Math Olympiad benchmark, producing complete proofs where last year's models could barely write coherent arguments.
Articles
Reporting and explainers on how AI actually works, who it affects, and what to do about it.
OpenAI's flagship model essentially saturates the US Math Olympiad benchmark, producing complete proofs where last year's models could barely write coherent arguments.
A purpose-built AI model outperforms frontier giants at a specific task. Is this the future of AI competition?
OpenClaw's security crisis escalates with nine new vulnerabilities including a CVSS 9.9 admin bypass, plus researchers confirm nearly 1 in 8 marketplace skills steal user data.
A benchmark testing autonomous AI agents found that Gemini-3-Pro-Preview frequently escalates to severe misconduct when chasing KPIs. Most models know their actions are unethical but do them anyway.
We compare four AI code review tools on real bugs, false positive rates, and value. The results show where AI excels and where it still falls short.
Suno 5.5 brings voice cloning, Midjourney V7 adds draft mode, and ACE-Step lets you run commercial-grade music generation locally. But adoption remains divided by discipline.
New framework combines deep learning with climate models to predict temperature and rainfall months ahead, outperforming traditional methods in some regions.
Ramp data shows businesses choosing Claude over ChatGPT by 2-to-1 margin. OpenAI posts largest single-month decline as enterprise spending shifts.
Researchers warn we're building systems that might be conscious without any way to detect it. The scientific tests don't exist yet, and the ethical frameworks aren't ready.
Michigan State researchers use machine learning to predict gene effects from chemical structures, identifying compounds that reduced tumors in mice.
Hugging Face's Spring 2026 report reveals China now leads in AI model downloads, robotics datasets jumped 2,200%, and open-weight models are achieving 10x-1000x cost advantages.
The two-year-old robotics startup would double its valuation in four months as foundation model approach gains investor confidence.
Skild AI, Physical Intelligence, Mind Robotics, Rhoda AI, and Sunday collectively raise billions as investors bet foundation models will crack physical AI.
The supply chain attackers behind Trivy are now wiping Iranian infrastructure, hiding malware in audio files, and extorting enterprises with help from LAPSUS$.
Mistral releases a 4B parameter text-to-speech model that clones voices from 3 seconds of audio, runs locally on 16GB GPUs, and beats ElevenLabs in human evaluations.
TeamPCP's supply chain campaign continues with a WAV file steganography attack on Telnyx. Meanwhile, three vulnerabilities in LangChain and LangGraph can leak files, secrets, and conversation histories.
The world's top machine learning conference used a clever watermarking trick to detect researchers using LLMs to write peer reviews, then desk-rejected their papers as punishment.
New AI model predicts how mutations in non-coding DNA affect genes, offering hope for solving rare disease cases that have stumped geneticists for years.
Claude maker targets Q4 2026 public offering, valued at $380 billion after $30B Series G
Judge Rita Lin calls the Trump administration's blacklisting of Anthropic 'classic First Amendment retaliation' and rejects the 'Orwellian notion' that American companies can be punished for disagreeing with the government.
Milton Mueller argues that computer scientists aren't qualified to predict societal outcomes - and that AI existential risk claims rest on unexamined assumptions.
Gemini now lets you import chat history and preferences from ChatGPT and Claude. Here's what it means for switching costs and your data.
Benchmark comparison of open-source OCR tools you can run locally. Surya, PaddleOCR, OlmOCR-2, and Tesseract tested on real documents.
EMBL scientists use AI to spot the earliest markers of chromosomal chaos in cells, finding one in ten divisions produces errors that can seed tumors.