I've been daily driving Pi for over a month now and never looking back. The latest crop of frontier models have really complex bash calls, so I've written an extension that rewrites them into plain English while still making them easy to reveal in full, unlike some harnesses which seem to try to hide this from the user which is pretty irresponsible IMO.[0]
Pi has been my goto agent for several months now, the project is great and basically runs everywhere. If you like Pi, I suggest taking a look at Deepseek Harness too.
For those using Claude, there's pi-claude-code[0] which is currently still the way to use Ant subs in Pi.
Is using Codex through Pi allowed? I've been a Claude user, but a lot of the things mentioned bother me as well. I see it constantly make mistakes due to (simple) shell aliases, etc..
Not sure I'll actually switch but I'm curious. I've thought about trying to build my own harness and use Codex usage to work through it without spending a fortune on credits.
Anthropic's subscription is the one that has historically been tough to use with other harnesses, where they have (had?) technical safeguards to ensure your subscription usage was limited to Claude Code CLI, with third party harnesses often implementing workarounds or faking it to use your Claude.ai subscription. Nowadays Jetbrains Air does officially allow using your Claude.ai subscription.
But OpenAI's subscription has seemingly always been available to use in other harnesses without trouble.
Yes, you can use your Codex subscription with Pi without any limitations. It supports Codex out of the box. You may need to install extensions for things like tracking usage and limit resets though. Pi only provides basic Codex integration.
> having a ton of CLAUDE.md files scattered about various GitHub hobby repos is great guerrilla marketing
Their real marketing comes/came from https://github.com/claude phenomenon. Irrespective of what underlying model you are using the claude code harness just sticks this github user as git commit author and then you have to make an effort to remove it. That is just odd.
I think Anthropic is working on allowing AGENTS.md files as well. However, Claude.md files take precedence if present. If not, you fall back to AGENTS.md.
Anecdotally I run semi nightly harness and local model test runs, and opencode consistently outperforms ohmypi on various code bench tasks.
Note though that this for local models that can run on a 5090, so, kinda dumb models with a small context window. Also they might perform differently on a custom pi setup rather than the very feature rich ohmypi.
Mostly the runs just cause me to get decision fatigue so I typically just use opencode and let a frontier model choose whatever cheaper models it wants to orchestrate with, so long as it's one that's not on the same token budget.
P.S. I'd always wondered who it was that got caleb.software before I could... Outshot in a western duel!
When you say "kinda dumb" it's worth noting that the smaller coding models tend not to actually need system prompts to be helpful coding assistants, etc. Qwen maybe more than Muse Glimmer? I've not tested Qwen 3.8 27B enough yet. Even Gemma 4 didn't seem to need much.
Muse Glimmer is so absolutely tuned for one particular pattern (turn-by-turn agentic coding and assistance) that it will do it more or less without a prompt at all (except to set the reasoning level maybe). Beyond the tools announcement, it feels to me like it almost only needs a prompt to do plan mode stuff. If LM Studio (in either of its forms) or Unsloth Desktop had proper support for running inside a linux container rather than just a filesystem sandbox, you'd barely need a dedicated coding agent at all. Osaurus gets even closer.
I wanted to like Opencode but I can't, really. I am enjoying Pi slightly more because it's just way less convoluted and obscure.
FWIW I use Paseo for the GUI, which to some extent eliminates that aspect of decision fatigue; if you like Paseo's UI well enough, you can have it with almost any harness. And since it can connect to its daemon via ssh you can still have the desktop GUI with a container. (Not clear to me if Opencode v2 Desktop makes remote connections to the daemon or just wraps the web interface like v1 did)
> Someone at Anthropic must’ve caught wind… just a day after making my decision, they announced Claude Code would finally start supporting AGENTS.md files in an effort to get me to stay.
Forgive me for asking if I should know, but who is the author? Is he famous?
>> Someone at Anthropic must’ve caught wind… just a day after making my decision, they announced Claude Code would finally start supporting AGENTS.md files in an effort to get me to stay.
> Forgive me for asking if I should know, but who is the author? Is he famous?
I think that was tongue in cheek. And by that I don't mean the author's tongue was actually in his cheek.
I also think it went over your head. And by that I don't mean I think it literally went over your head.
It's clearly a joke, but I can forgive some people for taking it literally, given the impact that certain well-known devs have on other companies' decision-making. It's a common enough line that it's plausible this could have been a VIP whose opinions swayed a company. Not him and not Anthropic, of course, but it does happen.
While I agree the sarcasm is obvious, important to remember sarcasm doesnt really translate well if english is not your first language, and notoriously does not translate well online regardless, see:
The whole AGENTS vs CLAUDE md issue was bound to be resolved. If the models are getting smarter and harnesses getting better - how could you have a product so dumb that it ignores whatever skills, instructions, references, other things that are part of the folder context it is mounted in.
Having claude code ignore agents.md just made it seem dumb. At this point, you probably don't need either.
> It seems like there isn’t someone high-up whose job is strictly to say “no” to these requests, which is incredibly important when it comes to product engineering
Ah, product management - we know so few of ye who know what you're doing.
Man, my M1 Pro unfortunately only has 16GB of RAM; it has been such a limiter to using local models. I'll give kapitsa a try, I see I should be able to squeeze some smaller models on
But this feels like six weeks (or more) out-of-date. And so it is - from mid-September. And I know that observing such a thing sounds bizarre and teenage, but that's kind of a long time right now; we're at a moment in history when things are moving fast.
We're not really fully ready for the traffic TBH, but for those who few of you choose to click, this is what we've been doing with pi lately:
[0] https://github.com/siraben/pi-tool-summaries
For those using Claude, there's pi-claude-code[0] which is currently still the way to use Ant subs in Pi.
[0] https://github.com/earendil-works/pi-claude-code
Not sure I'll actually switch but I'm curious. I've thought about trying to build my own harness and use Codex usage to work through it without spending a fortune on credits.
But OpenAI's subscription has seemingly always been available to use in other harnesses without trouble.
Pi/opencode apparently make 10% already of all traffic already:
https://manifest.build/blog/chatgpt-plus-tokens-third-party-...
Their real marketing comes/came from https://github.com/claude phenomenon. Irrespective of what underlying model you are using the claude code harness just sticks this github user as git commit author and then you have to make an effort to remove it. That is just odd.
Note though that this for local models that can run on a 5090, so, kinda dumb models with a small context window. Also they might perform differently on a custom pi setup rather than the very feature rich ohmypi.
Mostly the runs just cause me to get decision fatigue so I typically just use opencode and let a frontier model choose whatever cheaper models it wants to orchestrate with, so long as it's one that's not on the same token budget.
P.S. I'd always wondered who it was that got caleb.software before I could... Outshot in a western duel!
When you say "kinda dumb" it's worth noting that the smaller coding models tend not to actually need system prompts to be helpful coding assistants, etc. Qwen maybe more than Muse Glimmer? I've not tested Qwen 3.8 27B enough yet. Even Gemma 4 didn't seem to need much.
Muse Glimmer is so absolutely tuned for one particular pattern (turn-by-turn agentic coding and assistance) that it will do it more or less without a prompt at all (except to set the reasoning level maybe). Beyond the tools announcement, it feels to me like it almost only needs a prompt to do plan mode stuff. If LM Studio (in either of its forms) or Unsloth Desktop had proper support for running inside a linux container rather than just a filesystem sandbox, you'd barely need a dedicated coding agent at all. Osaurus gets even closer.
I wanted to like Opencode but I can't, really. I am enjoying Pi slightly more because it's just way less convoluted and obscure.
FWIW I use Paseo for the GUI, which to some extent eliminates that aspect of decision fatigue; if you like Paseo's UI well enough, you can have it with almost any harness. And since it can connect to its daemon via ssh you can still have the desktop GUI with a container. (Not clear to me if Opencode v2 Desktop makes remote connections to the daemon or just wraps the web interface like v1 did)
Forgive me for asking if I should know, but who is the author? Is he famous?
> Forgive me for asking if I should know, but who is the author? Is he famous?
I think that was tongue in cheek. And by that I don't mean the author's tongue was actually in his cheek.
I also think it went over your head. And by that I don't mean I think it literally went over your head.
https://en.wikipedia.org/wiki/Poe%27s_law
ooof, that is a lot!
And we're supposed to believe them and start one-shotting our projects...
Having claude code ignore agents.md just made it seem dumb. At this point, you probably don't need either.
There's only so much stuff I have the patience to tell it in a session before I write it down somewhere.
Ah, product management - we know so few of ye who know what you're doing.
try kapitsa as the inference engine with pi as the harness
It works really well
But this feels like six weeks (or more) out-of-date. And so it is - from mid-September. And I know that observing such a thing sounds bizarre and teenage, but that's kind of a long time right now; we're at a moment in history when things are moving fast.
We're not really fully ready for the traffic TBH, but for those who few of you choose to click, this is what we've been doing with pi lately:
https://magenta.cryptograss.live/