Keep dabit3/sonic-agent caught up with upstream NousResearch/hermes-agent every day.
If there are new upstream commits, produce ONE open PR into main that merges those
changes while preserving Sonic-specific branding and speed work. Verify with tests,
leave proof in the PR, self-review the PR, and stop so the human only has to review/merge
the morning PR.
If there is nothing new upstream, do NOT open a PR. Leave a short “no upstream changes” note and end the session.
Ramp published a post about Glass, the internal AI productivity suite they built for every employee. It's impressive work: auto-configured integrations, reusable skills, persistent memory, scheduled automations, Slack-native assistants, and 350+ skills reaching 99% adoption. Their post is worth reading in full.
If you want the same setup for your team, Devin supports these capabilities out of the box. No dedicated engineering team required. This post walks through each one and how to set it up.
Devin is an AI teammate that can write code and analyze data. Each session runs in its own cloud VM with a code editor, browser, terminal, and shell. Your repos are cloned, dependencies installed, and tools connected before Devin starts working. For data work, Devin includes a specialized Data Analyst agent (DANA) that queries
Devin can build, test, and iterate on Expo and React Native apps. This guide walks through how to configure Devin's environment so it's ready to work on your Expo project from the first session.
Expo provides a managed workflow that simplifies React Native development. For Devin, the key advantage is Expo's web preview — Devin can run your app in its built-in browser without needing a physical device or emulator. This means Devin can visually verify UI changes, test navigation flows, and debug layout issues during a session.
Ramp built Inspect, an incredible internal background coding agent responsible for ~30% of their merged PRs. Building it required months of infrastructure work: managed sandboxes, agent orchestration, session state, and custom tooling.
Cloud infrastructure like AWS and Cloudflare did away with the need to manage physical servers, patch operating systems, and maintain uptime. Instead of managing infrastructure, developers could focus on their application logic while the cloud provider handled the rest.
Devin applies the same principle to coding agents: the sandbox, dev environment, agent runtime, orchestration, integrations, and tooling are all fully managed. You bring the prompt; Devin handles everything behind it.
Here's how to replicate Inspect's functionality on top of the Devin API. You skip the months of infrastructure work and go straight to building the parts uniqu
PI is a TypeScript toolkit for building AI agents. It's a monorepo of packages that layer on top of each other: pi-ai handles LLM communication across providers, pi-agent-core adds the agent loop with tool calling, pi-coding-agent gives you a full coding agent with built-in tools, session persistence, and extensibility, and pi-tui provides a terminal UI for building CLI interfaces.
These are the same packages that power OpenClaw. This guide walks through each layer, progressively building up to a fully featured coding assistant with a terminal UI, session persistence, and custom tools.
By understanding how to compose these layers, you can build production-grade agentic software on your own terms, without being locked into a specific abstraction.
Pi was created by @badlogicgames. This is a great writeup from him that explains some of the design decisions made when creating it.
See more of my writing here. Also check out Devin
In this post, I'll start from scratch and build up to OpenClaw's architecture step by step, showing how you could have invented it yourself from first principles, using nothing but a messaging API, an LLM, and the desire to make AI actually useful outside the chat window.
End goal: understand how persistent AI assistants work, so you can build your own (or become an OpenClaw power user).
When you use ChatGPT or Claude in a browser, there are several limitations:
| #!/usr/bin/env python3 | |
| # mini-openclaw.py - A minimal OpenClaw clone | |
| # Run: uv run --with anthropic --with schedule python mini-openclaw.py | |
| import anthropic | |
| import subprocess | |
| import json | |
| import os | |
| import re | |
| import threading |
| import Anthropic from "@anthropic-ai/sdk"; | |
| import { Pool } from "pg"; | |
| const anthropic = new Anthropic(); | |
| // dn connection - Claude never sees this | |
| const db = new Pool({ | |
| host: "your-database-host.com", | |
| port: 5432, | |
| database: "sales_db", |
This is cross-posted from the original x post.
What makes Claude Code powerful is surprisingly simple: it's a loop that lets an AI read files, run commands, and iterate until a task is done.
The complexity comes from handling edge cases, building a good UX, and integrating with real development workflows.
In this post, I'll start from scratch and build up to Claude Code's architecture step by step, showing how you could have invented it yourself from first principles, using nothing but a terminal, an LLM API, and the desire to make AI actually useful.