Skip to main content

Research

Research Briefs

Each Research Brief covers one market shift and links to the company pages, compare pages, and methodology behind it.

Start with the latest brief, then use the archive when you want older context.

Explore the evidence

Open the company pages, compare pages, and methodology behind the latest brief.

Supporting archive

Older articles that still add context

Research Briefs are the active publication format. These older articles remain available when you need extra background.

Interface of Google's MIRAGE benchmark showing a multimodal AI's reasoning error analysis on a complex visual task.

Anthropic's Claude Sonnet Leads New Practical AI Benchmark

By Nick Allyn3 min read

A benchmark of 25 AI models across 125 real-world business tasks has put Anthropic 's Claude Sonnet at the top on output quality, while finding that OpenAI 's newer GPT-5 series is slower and no better than GPT-4.1. The analysis, published by entrepreneur Cristian Tala Sánchez ,...

Abstract editorial illustration of per-task AI pricing contrasted with traditional software subscriptions.

Anthropic AI Prices Labor, Not Software, in New Model

By Nick Allyn5 min read

Anthropic's new AI code review feature costs $15 to $25 per pull request. That price point, reported last week, set off an argument in developer communities that went well past sticker shock. The real friction wasn't the number itself. It was what the number implies about how AI...

Return hook

Keep up with the next Research Brief

A short weekly note with the newest brief, one meaningful market shift, and direct routes into the supporting evidence.