I've been running my own personal agent for months. Hermes on a server I control, reached through Telegram and a desktop app. So when Instinct and Muse launched, nothing they did was new to me. And I still wanted one. The capability isn't new. What's new is that the setup is gone. You text a number or download an app and it starts working. I'm fascinated by these UX paradigms: - A single thread is the right default. No project to pick, you just say the thing. - The agent runs the loop now. In my setup I write the monitoring myself. - Trust is the open question. The model sees one request at a time. The harness sees your whole life. These agents are built for productivity. I spend my days building agents that guide patients through their care, and healthcare is part of everyone's personal journey too. Different need, different UX, and launches like these teach us a lot. Full breakdown: hadijaveed.me/2026/09/22/per…

My Mac was heating up running multiple Claudes, Codex, and Hermes in parallel. So I moved my agents to a $90/mo Linux box in a data center. They run forever and my context compounds there. Full setup and how I secured it: hadijaveed.me/2026/08/27/age… my laptop stays cold now.

How are you running coding agents these days? I keep seeing two camps: 1. Claude Managed Agents or Claude on the web, where the agent works in a cloud sandbox and opens a PR on GitHub 2. Agents running directly on your own laptop The part I'm most curious about is testing. When the agent builds something like a web dashboard, where does it actually run? How do you check localhost things when the environment lives in someone else's cloud? I've been experimenting with my own setup on a dedicated server. Want to hear what others use before I share it. Still a lot to learn here. With Devin and all these agent factories popping up, it feels like everyone is answering the same question differently: where does the agent's computer live? If you're doing this daily, what does your setup look like

I'm usually the optimist in the room when it comes to AI. Lately I'm less sure. The promise was that AI would level the playing field. Anyone with curiosity and a laptop could compete with anyone. But watching the latest generation of frontier models roll out, I see a different pattern forming. The best intelligence is metered by API usage. The more you can pay, the more leverage you get. Companies and individuals with deep pockets compound their advantage while everyone else works with second-tier tools. If that holds, AI doesn't flatten the hierarchy. It hardens it. Some people could end up as a permanent underclass, priced out of the very intelligence that was supposed to lift them up. This is why open source matters more than ever. Not as a nice-to-have, but as the only real counterweight. I hope open models catch up fast, because the alternative is a world where access to intelligence is just another thing money buys. Just a thought I keep coming back to.

repo url: github.com/byaan-ai/byaan

Six months into using Claude Code full-time. My tmux splits have replaced every IDE panel I used to depend on. The context stays in the terminal, the history stays in git, and the AI sees exactly what I see. No GUI abstraction layer eating tokens.

Spider benchmark says text-to-SQL is 86% accurate. Spider 2.0, which uses real enterprise schemas, says 6%. That is not a rounding error. That is the difference between a demo and production. Most AI data tools are optimized for the demo.

this isn’t about AI replacing doctors. It’s about access. Patients are already using AI as their first touchpoint. Consumer-driven healthcare is here. Health systems need to meet patients where they are through technology

My Claude Code setup right now: export CLAUDE_CODE_NO_FLICKER=1 alias cc="~/.local/bin/claude --permission-mode auto" Two lines. Biggest productivity unlock I’ve had in months. What changed: No flicker modeL feels like a real app, not terminal spam Auto permissions: no more clicking “approve” 40 times Just give it a task: come back to a PR The key insight: Manual approvals aren’t safety. They’re just friction. Auto mode handles the risky stuff. Everything else moves. A few quick upgrades: Run /powerup (this is very new, to learn features) Add a CLAUDE.md (teaches it your stack + conventions) Create custom slash commands for repeat workflows or skills super helpful Try the CLI for a week. Same product… but the CLI UX just hits different. Feels faster, locks you in, and honestly way more fun with tmux.