We evaluate DeepCode on the PaperBench benchmark (released by OpenAI), a rigorous testbed requiring AI agents to independently reproduce 20 ICML 2024 papers from scratch. The benchmark comprises 8,316 ...

Abstract: Guessing Random Additive Noise Decoding (GRAND), a new algorithm of error correction code, has been proposed in recent years. The algorithm can correct the ...

On February 2nd, 2025, computer scientist and OpenAI co-founder Andrej Karpathy made a flippant tweet that launched a new phrase into the internet’s collective consciousness. He posted that he’d ...

Microsoft has released a new report showing what people used its AI assistant Copilot for in 2025. The analysis is based on 37.5 million de-identified conversations and shows that in addition to ...

The 300-person startup hopes bringing designers aboard will give it an edge in an increasingly competitive AI software market. Cursor, the wildly popular AI coding startup, is launching a new feature ...

The exhilarating speed of AI-assisted development must be united with a human mind that bridges inspiration and engineering. Without it, vibe coding becomes a fast track to crushing technical debt. If ...

On Tuesday, French AI startup Mistral AI released Devstral 2, a 123 billion parameter open-weights coding model designed to work as part of an autonomous software engineering agent. The model achieves ...

For the past several months, my social media feed has been flooded with people bragging about spinning up apps and websites over a weekend without any engineering help or coding — with just vibes.

Volvo Car AB is looking for partnerships for its new central software stack that’ll run on all of its future electric models, a sign the carmaker has overcome earlier coding glitches that delayed ...

Abstract: This letter investigates the achievable rate of multi-stream spatiotemporal channel coding (STCC-MS) with linear receivers. We first establish the transmission model of STCC-MS and explore ...

Building a golden path to AI Your team members may not be straight-up vibe coding, but they’re almost certainly using AI tools that management hasn’t signed off on, which is like shadow IT on steroids ...