Posted by eggbrain 1 day ago
Ask HN: What's your current AI agentic setup?
1) What AI subscriptions are you personally paying for? (Claude, ChatGPT/Codex, Cursor, etc)
- Which tiers?
- Have you purchased multiple subscriptions from the same provider (eg two Claude accounts)?
2) Are you using models outside (or instead) of those subscriptions? For example, pay-as-you-go APIs, OpenRouter, or locally hosted models.
- Which models and providers?
3) What coding harnesses and interfaces are you using? Are you sticking with the default CLIs or IDE integrations, or using alternatives like OpenCode, Pi, Aider, Muse Code, or custom-built harnesses?
- What made you switch harnesses (if you switched)?
4) Are you using any agent orchestration or swarm tools?
- For example, Munder Difflin, Gas Town, Ruflo, or something custom.
5) Are you using broader agent platforms like OpenClaw / Hermes Agent / Grok bot / Dots / Muse?
- If you are using them, what are your main use cases?
-----
I know my workflow has differed significantly from a year ago, and so it would be interesting to capture a snapshot today of what people are using and why.
I am salaried making > six figures and all I do is press the think button on the funny little screen.
I largely assume that all the micro optimizations people are making to their "agentic setups" is just bikeshedding. There's not much else for them to think about now. Probably makes people feel useful if they can get 0.5/1% improvements to their job-doer robot.
That being said, I've found peace in just embracing it. Work didn't require me to think anymore, so I've found some of that in trying to get very very good at a videogame. Quantifiable improvement, problem solving, adapting to new scenarios, etc.
But for work all that is needed is "computa, fix the bug".
Edit: To clarify, I spent my entire teens studying this field. Starting at age 11. Over 15 years of this being the sole focus in my life. I enjoyed the process more than anything in life. But that process is no longer what is being asked of us. I see no point in continuing to obsess over a field that, frankly, doesn't exist anymore. At least not in the shape that I have an ability to derive fulfillment from. If you still can, I'm both happy for you and envious of you.
I am not an AI bro. I don't like AI. It hasn't made my life better in any way shape or form. In fact, it has only made it worse.
Edit 2: I also learned how to read music, started playing the Piano, and started to learn how to properly draw. I don't find any of these things as fulfilling as I did programming pre-AI, but they aren't nothing.
I just run Claude/Codex raw, always did, before I was always forcing plan mode before any task, but they are so good right now that not even that is needed anymore 90% of the times.
Also before I was at least using terminals, now? Not even that, Claude Desktop and ChatGPT desktop apps all the way.
I would rather AI not exist. But it does. No sense in ignoring it.
I don't know what either of those two things actually mean.
I have so many questions:
* What kind of work are your agents producing?
* Why is your employer happy with that low quality output?
And most important,
* are you hiring? I could finally finish tetris!
- Because it solves the problems they need solved.
- Before AI companies whitelabelled humanity's accomplishments, I was very performance-minded. I still am I guess. I benchmark every function and if it's slower than I anticipated I tell the computer to fix it. A few months ago I would sometimes have to do it myself. Now I don't.
2. I have an OpenRouter $50 pay as you go API key that I use for Mercury 3 at this point, primarily for tokens per second. I also use GLM 5.3 when I inevitably get [cyber] blocked (or for reverse engineering in general, apparently) by the morality police at Anthropic.
3. I run things pretty stock. OpenCode with OpenRouter but otherwise I just use Claude code.
4. I've never used these, but I'm curious about them at this point.
5. Not yet, but also interested. My plan is to use Hermes for some personal agent work on a local model, but I have not set it up to test this yet.
I try to keep the harness setup simple and portable, I use Claude Code or Codex with their subscriptions, I invest a lot of time in my skills which are shared between them.
Disclaimer: I am the Paseo maintainer.
After around 30 days I’ve implemented around 90% of how you manage a team in emergency medicine by accident and then applied the rest of it to it.
I’m quite happy with what they produce.
1. I got a set of rules how it’s allowed to do work, and more importantly: what’s forbidden.
2. One state file per repo. It’s getting recorded what was tried, its results and the learnings from the experiments.
3. Tasks as Org mode files, one file per task and who works on it, aka codex, Claude code and which model.
Agent Hooks hold this together, the task and state files get updated automatically.
Code work happens on one branch per task. A second agent reviews each plan before code gets written.
I’m still baffled how well this works.