79
MENTIONS
21
EPISODES
21
PODCASTS
Search complete. 79 mentions across 21 episodes found for "Pi".
Oct 15, 2026
HackerNoon AINARRATOR
OpenCode, free client, some free models, Zen per token, or your own key.
HackerNoon AINARRATOR
Pi, free client, your own key.
HackerNoon AINARRATOR
DeepSeek, no subscription, V4 Pro, zero dollars.
HackerNoon AINARRATOR
Sixty-six, one dollar, ninety-eight per M, off-peak.
31 MINS LATER
HackerNoon AINARRATOR
Four, GitHub Copilot CLI.
HackerNoon AINARRATOR
Its in commands sit right next to your pull requests, which is a real advantage.
HackerNoon AINARRATOR
Five, Pi.
HackerNoon AINARRATOR
It is minimal, transparent, scriptable, and embeddable, which makes it the best foundation for building your own orchestrator.
LisaHOST
For those of you who are unaware of what Omnigent is, it is an meta harness over your other harnesses.
LisaHOST
So for instance, you want to use both, you know, Codex and Claude, for instance, or, you know, in my case, a lot of the open source, uh, harnesses as well, like, you know, Hermes, um, Pi, uh, and you're running other local models.
LisaHOST
Uh, to be able to have a layer over that where you can share policies across them, you can sandbox them, you can collaborate, and also do things like cross reviews.
LisaHOST
So let's say I make a PR with Claude, um, being able to automatically have Codex review that particular PR is incredibly useful and, uh, oftentimes increases the base quality of the code, uh, the majority of the time as well.
Hugo Bowne-AndersonHOST
Does that work?
Tomasz TunguzGUEST
Okay, so I live in Pi, and I run local models.
Tomasz TunguzGUEST
And I love Quinn35BA3B.
Tomasz TunguzGUEST
And I am going to run a skill in Pi, which is public company analysis for Figma.
Tomasz TunguzGUEST
And Figma just announced earnings today, and they crushed earnings.
Tomasz TunguzGUEST
Way to go, Dylan and team.
Tomasz TunguzGUEST
But right now, I think this is PresentJS, and this is the best library I've found for customization of software.
Hugo Bowne-AndersonHOST
Awesome.
swyxHOST
That's a good hint.
ParzivalHOST
[digital music] Welcome to The ASI Pill.
ParzivalHOST
Today, Alex Zhang, the MIT researcher behind recursive language models from his conversation with Swyx on the Latent Space Podcast, reshaped for The ASI Pilled.
ParzivalHOST
His claim is that the models are already geniuses, and we have been managing them all wrong.
ParzivalHOST
[digital sound] Alex Zhang is a second-year PhD student at MIT, and his paper on recursive language models has already been built into Prime Intellect's Agent, into Harvey's legal AI, and into the harnesses people threw at the Arc AGI3 benchmark.
28 MINS LATER
Alex ZhangGUEST
But there is, there is still kind of, uh, a lot of research to be done in this area around like, you know, what is necessary? Like what, what do we want to do? What design do we want? I mean, we probably don't want everything to be a swarm, but like where do we draw the line? Like can the agent design that or decide that for itself? Et cetera, et cetera.
Alex ZhangGUEST
So...
ParzivalHOST
That was the ASI Pill episode three hundred one with Alex Zhang from the Latent Space Podcast.
Fred WilsonGUEST
I mean, we had a moment on stage last week at the USB CEO Summit, where I was interviewing the CTO of Coinbase about some of the changes that they've made organizationally to be more AI-centric in how they do the work.
Fred WilsonGUEST
And he just casually mentioned that they moved away from Claude and they built their own harness, their own coding agent harness, on an open-source product called Pi.
Fred WilsonGUEST
Well, we're an investor in this company, Arendelle, who put Pi out into the market.
Fred WilsonGUEST
I didn't even know that this had happened.
Fred WilsonGUEST
I was just like, yes!
John McBrideHOST
I had sort of the opposite conclusion that like, I think you fell down kind of another tar pit that is like doing the next kind of like maxing of like AI things, and maybe it will get you just that little bit further.
John McBrideHOST
I think it was the same with like MCP, and then people had a bundle of MCPs, and it was like, "What am I doing with all this context?" You have like a harness with all these custom tools and like different ways it's orchestrating sub agent some things, and then you end up with like a gas town thing and you're just like, "Oh my God," like, [chuckles] "I can't manage this." There was a great post from the guys at Arandil who run Pi, and it was basically like all you need is a read tool, a write tool, and a bash tool, and the model can kinda take care of the rest.
John McBrideHOST
And I found that so interesting from a skills perspective as well, 'cause I've actually found myself like removing more skills as more capable models, even open source models that I'm running now kind of come down the pipeline.
John McBrideHOST
Like I, I'm running Qwen three eight Flash next, [chuckles] uh, four point floating point on these two sparks, and I'm actually shocked at how good it is at just like being able to write a little bit of bash one-liners.
Omar SantosGUEST
DJ Zampath, he actually owns all the AI platforms at Cisco.
Omar SantosGUEST
And me, we were kind of nerding out somewhat of solving what is actually going to be the SDLC, knowing that software factories are coming and more people are using, you know, whether it's Codex, Cloud Code, Pi, and OpenCode, and the plethora of agent harnesses plus plethora of models.
Omar SantosGUEST
what can we do that is more agnostic and harness agnostic that will actually bring the security best practice and then the reminders and hooks and rules and agent skills to that software factory for the next generation of the SDLC.
Omar SantosGUEST
So we called it Project CodeGuard internally, and it started with rules.
22 MINS LATER
Raj KrishnamurthyHOST
For a computer science student, right, who is coming out of college and who is trying to get into cybersecurity, And they're all very worried, right? What skills do they have to pick up today, in your opinion? How can they become successful, you know, becoming a cybersecurity engineer?
Omar SantosGUEST
Yeah, and from cybersecurity, you still need the fundamentals.
Omar SantosGUEST
You cannot just skip the fundamentals and say, I'm just going to learn how to manipulate open code or Pi or Hermes
Raj KrishnamurthyHOST
of security.
Chris AlexiukGUEST
If for no other reason, then you can generalize to all the changes they're gonna make.
Chris AlexiukGUEST
There's also, like, a very fun interplay where harnesses like to implement feature from other harnesses, right? So, like, when open code has a feature, like, Pi.dev will release a plugin for it or whatever like that.
Chris AlexiukGUEST
And so if we train a, a model that understands that open code feature well, when Pi.dev gets it, it understands it well there as well.
Chris AlexiukGUEST
So there, there, there's some of that.
Chris AlexiukGUEST
But, uh, okay, so that's the harness side.
denolfeHOST
Let's jump in.
denolfeHOST
Title, Pi 1.0, a stable, minimal, and extensible agent harness.
denolfeHOST
Source, Arendelle.
denolfeHOST
The Post announced the release of Pi 1.0, a simplified but flexible agent platform used globally.
denolfeHOST
This version added features like code mode for MCP support, virtual model extensions, deferred tool loading, cache warming, mid-conversation system messages, a new terminal user interface theme, and full-screen mode.
denolfeHOST
The developers emphasized that Pi remains true to its minimal roots, focusing on stability and proven features, allowing users to build long-running, durable agent applications without unnecessary complexity.
denolfeHOST
They also introduced PyDurable, an experimental branch for developing agents capable of handling long-duration tasks while maintaining minimalism and extensibility.
denolfeHOST
In the comments, the community showed mostly support, praising Pi for its simplicity, customizability, and stability.
Dave JonesHOST
Yes.
Dave JonesHOST
[laughs] I was, I was trying to get, I was trying to get Pi to go grab the MCP server docs while you were talking, but, uh-
Adam CurryHOST
Okay
Dave JonesHOST
...
Adam CurryHOST
People are always looking for something to do, something to listen to, or maybe something to watch.
Adam CurryHOST
And I have a feeling that this MCP server [clears throat] which I just call the podcast index robot, will help people discover things and will help podcasts be discovered for whatever reason.
Adam CurryHOST
So just if you've connected it to your, your, um, was it, uh, Pi?
Dave JonesHOST
Mm-hmm.
11 more episodes mention Pi.
Create an account to see the whole feed, search across every transcript, and follow the entities you care about.