Peter Gostev
PulseAugur coverage of Peter Gostev — every cluster mentioning Peter Gostev across labs, papers, and developer communities, ranked by signal.
-
CyberGym benchmark released to test AI agent capabilities
The CyberGym benchmark, designed to test AI agents' ability to perform complex tasks in simulated environments, has been released. This benchmark aims to evaluate the effectiveness of AI agents in dynamic and interactiv…
-
SQL controls Doom game mechanics in novel DOOMQL project
Peter Gostev has developed DOOMQL, a novel Doom-like game where SQL queries control game mechanics such as movement, collision, and combat, all within a SQLite database. The game is implemented as a Python terminal scri…
-
OpenAI launches GPT-5.6 and integrates Codex into ChatGPT Desktop
OpenAI has launched GPT-5.6, introducing three tiers: Sol, Terra, and Luna, with Sol's Ultra mode coordinating multiple agents for complex tasks. The company also rebranded its Codex product to ChatGPT Desktop and integ…