Guth Labs News
We are preparing a small, repeatable set of everyday task tests. Each future experiment fixes its agent instructions, tools, exact model and version, and task set.
Task set
Supplied fixture
Rewrite this three-sentence email in a friendly, clear tone: “Hi Sam. I need the document by Friday. Please send it when it is ready.”
Check criterion
Keeps the request and Friday deadline while using a friendly tone.
These starter tasks use sample data. Future results will show the agent setup, task outcome, time, steps, and cost.
Live trials are not open yet. Browse the tasks while we prepare the agent experiments.
Guth Labs News
Publisher updates about new models, alongside a small, fixed set of everyday task tests.
Publisher update
OpenAI News
GPT‑6 Astra improves Devin’s ability to test software and show that it works, with the goal of helping engineers review less code and ship more.
Read the publisher updatePublisher update
OpenAI News
GPT‑Live‑1 brings natural, full-duplex voice conversations to the API, with stronger instruction following, custom voices, and telephony support.
Read the publisher updatePublisher update
Hugging Face Blog
A Blog post by IBM Research on Hugging Face
Read the publisher updatePublisher update
OpenAI News
See how an MIT researcher uses GPT-5.6 Sol with Codex to autonomously run quantum computing experiments, analyze results, and calibrate qubits.
Read the publisher update