Continual Learning

Codex
Claude Code
OpenCode
Grok Build
Hermes
Pi
Oh My Pi
OpenClaw

Import agent activity

Choose what to import

Codex
Claude Code
OpenCode

Capture tool errors with Fix-it

Connect your harness

TypeScript · Python

Use the SDK in your app

Upload tasks and serve

Manage learning in Desktop

Download Desktop

1Ingest

typescript

import { openpond } from "./openpond.server";

const receipt = await openpond.addTask({
  id: eventId,
  name: "Support question",
  input: { question },
});

2Define

Run TimeTrainingAdapterGrader

3Serve

typescript

import { openpond } from "./openpond.server";

// Your accepted, active Model version.
const answer = await openpond.models.support.generate({
  prompt: question,
});
console.log(answer.text);

Ponder Training

Train with a workflow and adapter, then evaluate the candidate in an experiment.

Training workflows, adapters and their purpose
Workflow	Adapters	Purpose
SFT	LoRA Engram adapter Neuron training	Learn desired responses from reviewed demonstrations.
RL	LoRA Engram adapter Neuron training	Improve a policy using GRPO rewards.
Decision model	LoRA Engram adapter Neuron training	Train or score allowed choices from structured state.

Explore Ponder Training Explore continual learning Explore serverless inference

Example: RL + LoRA

Training loopLoRA adapter

RolloutsGenerate responses
GradeScore with your rewards
Update LoRATrain adapter weights
CheckpointVersion your adapter
EvaluateCompare with baseline
ServeCall your model API

New outcomes. Next adapter.

Models

Qwen3 0.6B
Qwen3 8B
Qwen3.5 4B
Qwen3.5 9B
Qwen3.8 27B
Nemotron Lightning 3.5 30B A3B

OpenPond Desktop & Harness

Work with agents beside your local projects, then turn a dataset into an evaluation plan without leaving the app.

Explore Desktop

Build your own models.