Fetching from the wire…
01
02
03
04
05
06
07
08
Research2026-08-18 · source-backed
MIT Technology Review covers a Stanford/MIT effort (Anka Reuel, Shayne Longpre) analyzing 24,521 donated conversations across 52 models from 2023-2025 against vendors' own usage reports. Because Anthropic filters for work-related use, nearly half of real conversations would be excluded. In the unfiltered set: harassment/hate at 27.5% versus Anthropic's reported 5.66%, sexual content 16.7% versus 2.4%, health/relationships 44.2% versus 31.2%. Anthropic's index rests on 1M conversations and OpenAI's on 1.5M, and neither is independently verifiable. If you're reasoning about AI adoption from vendor charts, the denominator is unknowable.
Each link below shares sources, entities, or timing with this story.
Moonshot AI released Kimi K3, a sparse mixture-of-experts activating 16 of 896 experts per token. That's about 1.8% of the pool live at any moment, with a 1M-token context window and native vision. Two new architectural pieces show up: Kimi Delta Attention and Attention Residu...
The beta workbench (macOS and Linux, Pro/Max/Team/Enterprise) acts as a project manager across 60-plus scientific databases, rendering 3D protein structures, genome browser tracks, and chemistry drawings alongside reproducible code. Anthropic explicitly said it is "not a new A...
Martin Alderson's essay "The upcoming AI margin collapse, part 1: GLM 5.2" hit 675 points and 462 comments on Hacker News, and it's the rare HN chart-topper that's actually about spreadsheet math instead of vibes. The argument is simple. Z.ai's GLM 5.2 delivers frontier-adjace...
Three frontier models shipped in a single week this month, and teams with a standing eval harness had a routing decision in hours. Anthropic's own agent-eval guidance says 20-50 tasks drawn from your real usage and real failures is enough to detect issues (DeepEval). DeepEval...
This Rust harness (+2,585 stars) competes on resource footprint rather than features: 27.8 MB PSS for a single session with local embedding disabled, claimed 13.9× less than Claude Code and 6× less than jcode's own embedding-enabled mode. Time-to-first-frame 14.0ms against a c...
This one's been building for days and it crystallized this week. Per The Register, the incident behind the US export-control block on Anthropic's Fable 5 and Mythos 5 wasn't a jailbreak or a guardrail bypass. It was a plain three-word prompt, "fix this code," run against CVE-l...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.