New playbook: when your own AI model pays. Read
Bridgewater Associates (investment management, ~1,000+ employees) reported this on 30 Jun 2026 in company self-report (source). It belongs to the learning loop layer.
A Qwen3-235B model fine-tuned on Tinker with investor-labelled data to do six daily information-filtering tasks (e.g. which news matters to a macro investor).
Accuracy 84.7% vs 78.2% for the best frontier model (29.8% fewer mistakes) at 13.8x lower inference cost per task; frontier models averaged ~50% with simple prompts and stayed under 80% with expert prompts.
The number covers six internal tasks on a held-out test set; Bridgewater's own evaluation.
“An explicit prompt can only convey the intuition an expert is able to put into words,” Bridgewater Associates said.
Reviewed corrections are kept as test cases and later used to train models.
Read the guide to the learning loop.
96%
lower estimated serving cost for one agent
A daily continual-learning loop for Sidekick's GraphQL agent: an expert-calibrated LLM judge scores production conversations, a self-healing pipeline repairs failures, and the repaired trajectories are trained into a smaller model (SFT, then GRPO against the judge) that replaces the frontier model (source).
1% → 99%
valid structured output after fine-tuning a small model
A proof of value on Red Hat OpenShift AI: GPT-4.1 generated synthetic training data, then small models (Phi-4 Mini 3.8B, Qwen 2.5 1.5B/7B) were fine-tuned for multilingual structured extraction from social media, deployable locally (source).
~7x
faster responses from a post-trained open model
A post-trained NVIDIA Nemotron 3.5 Lightning (30B MoE, 3B active) 'task expert' that took over Sidekick's GraphQL sub-agent from a frontier model, served on Shopify's own GPUs (source).