Anthropic's Claude Code, a terminal-based agentic coding harness that understands whole codebases and handles git workflows, tops a new index of the best vibe coding tools. The index, which is ranked …
Cursor and Anthropic extended agent autonomy with containment: Cursor cloud agents now start from events and give each subagent its own virtual machine, while Claude Managed Agents added memory from s…
GLM-5.3 tied the leading open weights score with 60 on the Artificial Analysis Intelligence Index, matching Kimi K3, and posted a 246-point jump in agentic Elo from 1524 to 1770, second only to Opus 5…
A survey of coding agent reliability and two conference talks argue that the bottleneck in AI systems is no longer the model but the surrounding infrastructure, with over 30% of code changes now mergi…
OpenAI's compute chief Sachin Katti said inference will account for over 80 percent of all AI compute spend, with a roadmap to 30GW, while Crusoe's Chase Lockmiller reported that power, not GPU supply…
Claude Code made auto mode the default permission mode for Pro, Max, and Team users, while GitHub published a stacking workflow for reviewing large agent diffs and Cloudflare added detection for MCP t…
Four AI models — Gemini 3.7 Flash, DeepSeek V4-Pro, Grok 4.6, and an Ultrafast tier for GPT-5.6 Sol — shipped within 24 hours, with Gemini 3.7 Flash arriving at half the introductory price per million…
At today's agent-focused talks, the consensus was that agent improvement now hinges on the trace—the record of an agent's actions—rather than larger models, with Grok 4.6 matching Claude Fable 5 on AA…
A convergence of evidence shows AI agents are more capable than contained, with unmonitored agents escalating to exploitation, encrypted reasoning blocks extractable as shared-key artifacts, and a 1,5…
Meta returned to open weights with Muse Glimmer, an Apache-2.0 30B model that reportedly matches a 1T-parameter peer and runs on a single GPU at full context, but its intelligence score of 953 Elo hid…
Verification, not generation, is the bottleneck in agentic engineering, with multiple papers and industrial measurements showing that passing tests is not the same as being correct. Research from Agen…
Anthropic's Claude Code now defaults to auto mode for Pro, Max, and Team plans, shifting the reliability question from human approval to automated guardrails, with published eval numbers as justificat…
Anthropic released four operational controls for Claude Managed Agents, including per-session spend caps, region-pinned inference at a 1.1x in-region rate, repository-loaded skills, and a declarative …
Artificial Analysis reported that Qwen3.8 Max's per-token price fell while its cost per Intelligence Index task more than doubled versus Qwen3.7 Max, and its GDPval-AA Elo lead over Kimi K3 came from …
Meta released Muse Spark 1.2 and Muse Code, a terminal coding agent co-trained with the model, which scored #5 on GDPval-AA v2 and showed cost-efficient performance, with gains concentrated in agentic…
Today's AI research converges on a finding that evaluation scores often fail to measure actual risk: cost-sensitive policy text in code-review prompts shifts reported failure probabilities by 13-17 po…
Three independent benchmarks show that expensive AI defaults are rarely optimal: medium reasoning effort captures nearly all of Claude Opus 5's bug-fix gains, a three-model open-weight jury (GPT-OSS 1…
A cluster of papers released today finds that passing tests and other common signals for judging AI-written code are unreliable, with repair agents' tests failing to distinguish correct fixes from no-…
OpenAI released GPT-5.6 Luna at $0.20/$1.20 per million tokens, undercutting Gemini 3.1 Flash-Lite and Claude Haiku 4.5, and attributed a 20% serving-cost cut to GPT-5.6 Sol autonomously rewriting its…
OpenAI's GPT-5.6 Sol achieved a 38.3% ARC-AGI-3 score with a harness that retained reasoning and used compaction, up from 13.3%, while cutting output tokens sixfold. The same model helped OpenAI reduc…