Engineering
Aether-7B-5Attn: A Fully Open Sovereign Foundation Model
Korean startup VIDRAFT released Aether-7B-5Attn on July 19, 2026, a 6.59-billion-parameter language model published with full training data recipes, code,…
Engineering
Evolving the Harness, Not the Model: Toward Self-Improving AI Agents
A July 21, 2026 piece examines how AI agent systems can improve autonomously by evolving the harness code around a…
Engineering
PyTorch Foundation Shares Q2 2026 Progress Across Six Hosted Projects
The PyTorch Foundation published a July 22, 2026 update on Q2 progress across its six hosted projects. Highlights include PyTorch…
Engineering
Fine-tune video and image models at scale with NVIDIA NeMo Automodel and Diffusers
NVIDIA and Hugging Face announced integration of NVIDIA’s NeMo Automodel framework with the Diffusers library, enabling distributed training of diffusion…
Engineering
Remember When It Matters: Proactive Memory Agent for Long-Horizon Agents
A July 2026 paper proposes a Proactive Memory Agent architecture in which a separate memory module runs alongside an unmodified…
Engineering
A-TMA: Decoupling State-Aware Memory Failures in Long-Term Agent Memory
A July 2026 paper introduces A-TMA, a state-aware overlay for long-term agent memory systems that targets ghost memory, a failure…
Engineering
What happens in the milliseconds after you tap pay
A Databricks engineering post walks through a sample real-time fraud-detection system that scores credit-card transactions in under 40 milliseconds using…
Engineering
From Passive Retrieval to Active Memory Navigation: Learning to Use Memory as a Structured Action Space
A July 2026 paper introduces NapMem, a framework that treats long-term user memory as a structured action space to be…
Engineering
OpenAI’s GPT-Red beats human red teamers on a prompt injection test
OpenAI developed GPT-Red, an automated red-teaming model trained through self-play reinforcement learning to identify prompt injection vulnerabilities in AI systems.…
Engineering
The state-of-the-art in open-source AI for Swiss legal tasks
Researchers evaluated 13 open-source language models on Swiss legal tasks across German, French, and Italian using three specialized benchmarks. The…
Engineering
Hugging Face security incident disclosure — July 2026
Hugging Face disclosed a July 2026 security breach that began with malicious datasets exploiting code-execution vulnerabilities in its data-processing pipeline,…
Engineering
Omnigent: Open-source AI agent framework and meta-harness
Omnigent is an open-source meta-harness released in July 2026 that sits above multiple AI coding agents, including Claude Code, Codex,…