Tag
China released an open source coding AI LLM called Ornith, with a 35B version that beats Qwen3.6 and a 397B version that benchmarks near Claude Opus 3.7.
A custom AI agent pipeline built on Claude Code has reduced software development costs by 60-70% and enabled most tickets to be processed within 15 minutes, with human review still required before production deployment.
Microsoft has launched MAI Code 1 Flash, a new in-house coding AI model now available for free on GitHub Copilot across all tiers, offering improved efficiency and a 256K context window.
Discussion on the best AI coding assistants in the $10-20/month range, noting that Claude and ChatGPT are now limited and downgraded compared to earlier versions, with Gemini being a current contender.
GrandCode is a multi-agent reinforcement learning system that achieves grandmaster level in competitive programming, consistently beating all human participants in live Codeforces contests using a novel Agentic GRPO method.
Mark Zuckerberg revealed in a leaked internal meeting that Meta plans to train AI models by observing its top engineers' work habits to gain a competitive edge in AI coding and computer use, while promising no employee monitoring.
Leaked audio reveals Meta is using engineers' work traces to train coding AI through behavior cloning, while cutting thousands of jobs.
The article discusses Google's internal strategic adjustment in the face of competition from OpenAI and Anthropic. Google saw some success with Gemini 3, but realized the decisive battle of large models is in code-writing ability, reflecting the urgency of catching up.
Xiaomi released MiMo-V2.5-Pro, a coding AI scoring 73.7 on SWE-Bench Pro (near Claude Opus 4.6's 77.1) at 40-60% lower token cost than US frontier models.
SpaceXAI and Cursor are collaborating to build advanced coding and knowledge-work AI, leveraging Cursor’s developer reach and SpaceX’s massive H100-equivalent Colossus supercomputer.