Tops GPTs 
On August 24, 2026, OpenAI announced the GPT‑5.6 model family is available in Kiro, a development agent for AI-native coding. OpenAI and AWS say the models improve code quality and cost-efficiency in long-running tasks, with testing showing an ~82% cost reduction for Terra.
U.S. courts are issuing mixed rulings on whether training AI on copyrighted books is lawful. Judge William Alsup found training lawful but ordered Anthropic to pay $1.5 billion for using pirated sources, while other cases test fair use and authorship rules.
Berd, launched in 2026, is a desktop app aimed at engineers and designers who want a persistent AI agent inside a local workspace. The Open Source Mac app links projects, skills, extensions, automations, model providers and local context.
NVIDIA teams use ChatGPT Work to automate recurring tasks, connect external AI developments to internal priorities, and share reusable workflows across regions. During a 12-week GTC cycle, one workflow saved about 16 hours per week for a Go-To-Market strategist.
Researchers at HumeAI report that some open-source ASR models reproduce benchmark-specific transcripts even when the audio contradicts them. Tests on VoxPopuli and LibriSpeech reveal dataset-linked acoustic cues, masked-number recovery, and orthographic switching.
Ramp has launched Router, an API-based model routing service that lets users switch among LLMs from OpenAI, Anthropic, xAI and others. The service is US-only, free through the end of 2026 with a $26 credit, and continues to charge for model inference.
Inherent, a London lab founded by DeepMind alumni, says its agent Faraday replicated published research more effectively than Anthropic’s Claude Opus 4.8 and OpenAI’s GPT-5.5 while running on a 27B-parameter Qwen 3.6 model. The company highlights reward-based training and a focus on “research taste.
Meta is rolling out Pocket across the U.S., an app that uses AI prompts to generate small interactive games called “gizmos.” The experiences can use touch, tilt, photos and music; Pocket is based on an acqui-hire of the Gizmo team and replaces the Atma Sciences app.
Liquid AI released DSpark draft checkpoints for LFM2.5-1.2B-Instruct, LFM2.5-2.6B and LFM2.5-8B-A1B, reporting up to 3.18× GPU and 2.87× on-device throughput. The approach uses a small draft model and speculative decoding; checkpoints are available in Safetensors and GGUF.
Ramp data from more than 70,000 U.S. businesses shows Anthropic held about 44% market share in July versus OpenAI’s nearly 40%. Ramp economist Ara Kharazian says OpenAI is growing faster in Q3 to date, though the sample excludes large enterprises.