Fancy harnesses make coding agents more expensive but not better, study finds
A controlled study shows that switching between coding-agent harnesses like Claude Code, OpenCode, or mini-SWE-agent barely moves pass rates more than rerunning the same setup. Cost per task, however, can triple, driven mainly by the system prompt that gets resent with every…
AI Summary & Key Takeaways
- A controlled study shows that switching between coding-agent harnesses like Claude Code, OpenCode, or mini-SWE-agent barely moves pass rates more than rerunning the same setup
- Cost per task, however, can triple, driven mainly by the system prompt that gets resent with every…
Fancy harnesses make coding agents more expensive but not better, study finds
Related stories
Kimi K3.1 Pro approaches GPT-6 Astra and Claude Opus 5.5 — but costs significantly less
⚡️ Kimi K3.1 Pro approaches GPT-6 Astra and Claude Opus 5.5 — but costs significantly less. Chinese Moonshot AI unveiled a new version of its flagship model with a 1 million token context window. According to the company's internal tests, Kimi K3.1 Pro shows…
Russia plans to allocate 500 million rubles annually for AI and cybersecurity textbooks
📚 Russia plans to allocate 500 million rubles per year for AI and cybersecurity textbooks. The Russian Ministry of Education plans to annually direct 500 million rubles to provide students in grades 5–11 with textbooks on artificial intelligence and information security. In schools…
ElevenLabs unveils Chimera — a tool for creating realistic AI characters
🤖 ElevenLabs introduced Chimera — a tool for creating realistic AI characters. It combines the Eleven4 voice model and the Seedance 2.5 video generator. What Chimera can do: — create a character from a photo or image; — animate facial expressions; — voice text with a natural-sounding voice.
Anthropic and Theorem Used Claude for Mathematical Verification of 123 Open-Source Libraries
📌 Anthropic and Theorem used Claude for mathematical verification of 123 open-source libraries. Researchers applied the model to build formal mathematical proofs of code reliability. The basis was ready-made fuzzing tests from OSS-Fuzz, Google's service for…