> Canonical page: https://posttrainllm.com/inspiration/weco

A note of thanks · Research loops

# Weco AI

Weco's traceable, steerable autoresearch loop is a candidate to study when we have a concrete target, frozen evaluation, and fixed cost budget. It is on our radar, not installed in the factory.

01 · The idea

## What stayed with us

Weco's traceable, steerable autoresearch loop is a candidate to study when we have a concrete target, frozen evaluation, and fixed cost budget. It is on our radar, not installed in the factory.

02 · The local translation

## What we did with it

The appealing part of Weco is an inspectable search tree: a proposed code change, its run, the metric, and the next branch stay visible. That matches our preference for accountable attempts and a stable evaluator before a candidate is called better.

03 · The boundary

## Where the comparison stops

We have not run Weco here. The owner recorded a trial as conditional on a concrete target, frozen eval, held-out gate, regression metric, and fixed cost budget. Its product claims are not local results.

Source trail

## Follow the work

- Original project [Weco AI ↗](https://www.weco.ai/platform)
- Our context [Our recorded factory stance ↗](https://github.com/PostTrainLLM/posttrainllm/blob/main/AGENTS.md#external-autoresearch-products-to-retain)

Independent appreciation. The named projects have not endorsed or affiliated with PostTrainLLM.