Morning Edition · Thursday, July 9, 2026
Tech
xAI Ships Grok 4.5, a Cursor-Trained Coding Model Priced Below Rivals
The model resolves SWE-Bench Pro tasks with about a quarter of the output tokens Anthropic's Opus 4.8 uses, even as it scores lower on the benchmark itself.
Tech
Anthropic's Claude Sonnet 5 Pushes Agent-Grade Coding Into a Cheaper Tier
The mid-tier model scores 63.2 percent on SWE-Bench Pro and 80.4 percent on Terminal-Bench 2.1, narrowing the distance to the pricier Opus line.
Tech
OpenAI Retracts Its Endorsement of SWE-Bench Pro, Citing Broken Tasks
An audit of the 731-task public split flagged 27 percent of tasks as broken by an automated pipeline and 34 percent by human reviewers.
Tech
NVIDIA Says Nemotron 3 Ultra Leads Open Models on LangChain's Agent Suite
Tuned inside LangChain's Deep Agents harness, the model scores an aggregate 0.86 at $4.48 per run against $43.48 for the closest closed model.
Tech
Meta's Brain2Qwerty Decodes Typed Sentences From Brain Scans Without Surgery
The updated non-invasive pipeline reaches 61 percent average word accuracy from magnetoencephalography, up from about 8 percent for prior non-surgical methods.
Tech
OpenAI Launches GPT-Live, a Full-Duplex Voice Model for ChatGPT
Two models, GPT-Live-1 and a mini variant, listen and speak simultaneously and delegate hard questions to a frontier model in the background.
Tech
New Work Sharpens Memory Poisoning as a Standing Threat to LLM Agents
Persistent agent memory lets an attacker plant instructions through ordinary queries that direct future actions, with prior attacks reporting injection success above 90 percent under lab conditions.
Tech
Study Finds Trusted AI Sabotage Monitors Fail to Transfer Across Model Families
Monitors tuned against one or two untrusted models overfit to that lineage's calibration and lose reliability on models they were not tested against.
Tech
TriRoute Proposes One Learned Router for Attention, Experts, and KV-Cache
The method jointly allocates sparse computation across three axes that prior techniques each optimized in isolation.
Geopolitics
OpenAI Publishes National-Security Principles as Government Deals Expand
The framework accompanies cyber-defense partnerships with the United States and allied governments and pledges no domestic surveillance of U.S. persons.
Tech
Cloud and AI Vendors Flood Startups With Compute Credits to Lock Them In
Infrastructure grants now reach into the millions per startup, structured to make later migration to a competitor costly.
Tech
General Intuition, Kyutai, and Epic Open-Source a Playable 5B World Model
MIRA generates Rocket League 2v2 matches frame by frame at 20 frames per second on a single GPU, with dataset and code released.