Its really fantastic. I can't imagine why you'd go through the effort using Claude Code with other models when pi is a much better harness. There's tons of extensions already available, and its trivial to prompt an LLM to create your any new extension you want. Lacking creativity and want something from another harness?
> Run <other harness> in tmux and interrogate it how feature X works, then build me the equivalent as a pi extension.
Maybe in a few years there will be obvious patterns with harnesses having built really optimal flows, but right now it works so much better to experiment and try new approaches and prompts and flows, and pi is the easiest one to tweak and make it your own.
> but right now it works so much better to experiment and try new approaches and prompts and flows, and pi is the easiest one to tweak and make it your own.
That’s what really appeals to me. I’ve been fighting Claude Code’s attempts to put everything in memory lately (which is fine for personal preferences), when I prefer the repo to contain all the actual knowledge and learnings. Made me realise how these micro-improvements could ultimately, some day, lead to lock-in.
> Run <other harness> in tmux and interrogate it how feature X works, then build me the equivalent as a pi extension.
If you ever want to use other models, pi can do that. In the middle of a session I might switch from gpt-5.2 to opus and get it to do something or review something and then switch back to gpt. Since models are being released every few weeks this is interesting to compare models without having to switch to a different harness.
And if there’s any feature codex has that you want, just have pi run codex in a tmux session and interrogate it how said feature works, and recreate it in pi.
Interesting point. And from the agents point of view, it’s always joining at the last minute, and doesn’t stick around longer than its context window. There’s a lesson in there maybe…
The context window is the onboarding period. Every invocation is a new hire reading the codebase for the first time.
This is why architecture legibility keeps getting more important. Clean interfaces, small modules, good naming. Not because the human needs it (they already know the codebase) but because the agent has to reconstruct understanding from scratch every single time.
Brooks was right that the conceptual structure is the hard part. We just never had to make it this explicit before.
Unifi makes a doorbell and consumer (and commerical) security cameras which run and store data on a local device, but still reachable online with their app connecting directly to your device. I used their dream machine pro with a big HDD, but they're released a few other devices in the last few years which might be cheaper and use SSDs. And I think you could run the stack in docker. But if you want to hack it yourself, there's probably easier projects. If you want to spend a bit more but have everything more or less just work with nice hardware and apps, Ubiquity's Unifi system is really great for home security. Not to mention the wifi and other networking solutions they have.
Yeah, I think there is some confusion in this thread. eu-inc.org isn't an official source or anything, though von der Leyen did say "EU Inc" in her speech at Davos. The European Parliament specifically mentioned not liking "incorporated" because it was American terminology preferring "Societas Europaea Unificata" instead which is pretty funny.
Reminds me of all the official sounding websites with short domain names purporting to explain the purpose and mechanisms of the GDPR, which were almost all operated by advertising companies or industry groups representing them.
Honestly very confused by the people happy or agreeing with Anthropic here. You can use their API on a pay-per-use basis, or (as I interpreted the agreement) you can prepay as a subscription and use their service with hourly & weekly session limits.
What's changed is that I thought I was subscribing to use their API services, claude code as a service. They are now pushing it more as using only their specific CLI tool.
As a user, I am surprised, because why should it matter to them whether I open my terminal and start up using `claude code`, `opencode`, `pi`, or any other local client I want to send bits to their server.
Now, having done some work with other clients, I can kind of see the point of this change (to play devils' advocate): their subscription limits likely assume aggregate usage among all users doing X amount of coding, which when used with their own cli tool for coding works especially well with client side and service caching and tool-calls log filtering— something 3rd party clients also do to varying effectivness.
So I can imagine a reason why they might make this change, but again, I thought I was subscribing to a prepaid account where I can use their service within certain session limits, and I see no reason why the cli tool on my laptop would matter then.
Seems like a great day for Pulsar / StreamNative and Redpanda who are going to get a lot of new customers in the coming years.
When looking for alternatives to Kafka these are the most promising options I found, not counting RabbitMQ which needs no introduction.
Pulsar seems to be 'kafka done better'. The version of Kafka that Confluent used internally (Kora) seems closer to Pular as well. Pulsar has a lot of features that Kafka doesn't have, like per-message acknowledgement, similar to RabbitMQ. And it has protocol support for RabbitMQ and Kafka, so can be a drop-in replacement.
Redpanda seems like a great re-implementation of Kafka.
I'm hoping this boosts Pulsar's status and helps get some traction for StreamNative. It seems like the best technical solution for events and messaging. It just needs a bit more market adoption to make an easier choice for enterprise, in my opinion. This might be that moment.
> Run <other harness> in tmux and interrogate it how feature X works, then build me the equivalent as a pi extension.
Maybe in a few years there will be obvious patterns with harnesses having built really optimal flows, but right now it works so much better to experiment and try new approaches and prompts and flows, and pi is the easiest one to tweak and make it your own.
reply