Saltar al contenido principal

Claude Opus 4.8 beats GPT-5.5 on agentic coding. The dev take.

29 May 2026
2:26
2 reproducciones

May 29, 2026 — Three quick AI updates for engineers who ship. 1. Claude Opus 4.8 — beats GPT-5.5 on agentic coding (69.2% vs 58.6% on Anthropic's own bench), adds effort controls and Dynamic Workflows (hundreds of parallel subagents in Claude Code). Pricing holds at 5/25 USD per million tokens. 2. Anthropic's free Claude Code security plugin — flags 25 vulnerability patterns (eval, os.system, innerHTML and more) as you edit, with zero token cost. 157K installs in 24h. Install from /plugins (needs Claude Code 2.1.144+ and Python 3.8). 3. Gemini Interactions API breaks June 8 — legacy schema removed. The outputs array becomes a steps array and response_mime_type is dropped for a polymorphic response_format. Upgrade to SDK 2.0 and update how you read responses. 🔗 Sources: - Claude Opus 4.8: https://www.techtimes.com/articles/317349/20260528/claude-opus-48-remote-execution-leaves-four-times-fewer-code-flaws-unflagged-beats-gpt-55-coding.htm - Claude Code security plugin: https://www.helpnetsecurity.com/2026/05/27/anthropic-claude-code-security-guidance-plugin/ - Gemini Interactions API: https://ai.google.dev/gemini-api/docs/interactions-breaking-changes-may-2026 #Shorts #AINews #ClaudeOpus #ClaudeCode #GeminiAPI #AICoding #DevTools #Anthropic #OpenAI #SoftwareEngineering

Comentarios
Debes iniciar sesión para comentar.

No hay comentarios aún. ¡Sé el primero en comentar!