Saturday, April 25, 2026
8 stories covered by gentic.news intelligence
OpenAI launched GPT-5.5, an agentic model that tops Terminal-Bench 2.0 at 82.7% and surpasses Claude Opus 4.7 and Gemini 3.1 Pro on coding and math. However, independent testing shows higher hallucination rates and effective API costs 20% above GPT-5.4 despite doubled token prices.
Nvidia demonstrated training a billion-parameter language model using zero gradients or backpropagation, eliminating FP32 weights entirely. This could dramatically reduce memory and compute costs for LLM training.
A detailed technical guide argues that the model is not the hard part of building AI agents. The six-component harness — context management, memory, tools, control flow, verification, and coordination — is what separates production-grade agents from those that fail silently.
A tweet reports that researchers fed a language model a DNA sequence and asked it to generate a new virus, which it did. This highlights both the power and risk of generative AI in synthetic biology.
Meta's new paper reveals that coding agents with summary-based history reuse outperform those using raw logs, improving efficiency and success on complex tasks.
Future AGI open-sourced a full platform that aims to eliminate silent hallucination in production AI agents, offering runtime monitoring and intervention tools.
.png&w=1920&q=75)
DigitalOcean's paper introduces lightweight behavioral signals to rank 80k agent-user trajectories, achieving 82% informativeness in sampled reviews compared to 54% for random sampling, with no LLM overhead.
Researchers sent identical resumes to an AI hiring tool, changing only the name. One version was rejected, revealing systemic bias in automated hiring systems.