# Our research on the Construct Computer blog

> Every post on Our research from the Construct Computer blog, newest first.

- [Which AI employees do ChatGPT, Google and Claude recommend?](https://construct.computer/blog/ai-search-recommends-ai-employees/) (Nov 13, 2026): We asked six AI engines 60 buyer questions about AI employees: Lindy was named most, vendors' own lists were cited most, and we came up 3 times in 219.
- [Best cheap LLM for AI agents: 11 models through our agent loop](https://construct.computer/blog/cheap-llm-for-ai-agents/) (Nov 10, 2026): 11 cheap model setups on 11 real agent tasks: GLM-5.3-Flash and Gemini 3.8 Flash passed 19 of 22, and GLM cost least per passed task by far.
- [Clef vs Jev: a benchmark on five real AI agent decisions](https://construct.computer/blog/clef-vs-jev-benchmark/) (Oct 5, 2026): Cloudflare's Clef against TypeSafe's Jev on 117 labelled agent decisions: Clef 92.3%, Jev 88.6%, Clef-flash 82.1%, with drift, repeatability and cost.

---

The Markdown copy of https://construct.computer/blog/tag/research/. Everything Construct Computer publishes is listed in https://construct.computer/llms.txt.
