Engineering
AutoDesign: Meta-Harness Optimization for Long-Horizon Agentic Design
AutoDesign targets systems that turn multimodal content into structured design outputs, arguing that most such pipelines stay static rather than…
Engineering
Spatial Memory Agent: Experience-Grounded Procedure Memory for Spatial Intelligence
This paper proposes the Spatial Memory Agent (SMA), a framework that improves spatial reasoning in frozen vision-language models without any…
Engineering
Massive Activations in Hybrid Linear Attention Large Language Models
This research examines internal activation patterns in hybrid linear-attention language models and identifies two distinct phenomena: sharp pre-attention spikes that…
Engineering
Full-Bandwidth Transformer
The paper identifies a bottleneck in autoregressive transformers where only sampled output tokens are fed back into the model, discarding…
Opinions
Christian Science Monitor: Schools’ AI “Gold Rush” Is Outpacing the Evidence
The Christian Science Monitor reports that the K-12 education technology market built around generative AI has grown to roughly $730…
News
OpenAI to Introduce Ads on ChatGPT’s Free and Go Tiers in Europe This Month
OpenAI announced on August 15, 2026 that advertisements will begin appearing on ChatGPT’s Free and Go plans for users in…
News
Anthropic Projects First-Ever Quarterly Operating Profit on $10.9 Billion in Q2 Revenue
Anthropic told investors it is on track to report $10.9 billion in second-quarter 2026 revenue, up 127% from $4.8 billion…
News
Manus to Resume Independent Operations as Meta’s $2 Billion Acquisition Unwinds
AI agent startup Manus announced it will soon resume operating as an independent company as Meta’s roughly $2 billion acquisition…
News
OpenAI Pauses Parts of Astra Development Over Critical Cyber Capability Concerns
OpenAI has paused certain internal work on its upcoming Astra model after preliminary evaluations found the system’s agentic coding and…
News
OpenAI Previews Ultrafast Mode, Running GPT-5.6 Sol at 14x Speed on Cerebras Chips
OpenAI introduced a preview of “Ultrafast,” a new API service tier that runs its GPT-5.6 Sol model on Cerebras hardware…
News
Google Launches Gemini 3.7 Flash With Steep Price Cut for Coding and Agent Workloads
Google unveiled Gemini 3.7 Flash, a faster, cheaper multimodal model built for coding and AI agent workflows, arriving three weeks…
Engineering
Thought-Level Beam Search for Reasoning
This paper targets inefficient test-time compute scaling in large reasoning models, where independent sampling wastes resources and naive pruning fails…