Thomas D retweetledi
Thomas D
5.6K posts
The comparison is because gstack is the opposite of Smithers.
Smithers says “here is a powerful way to use typescript to orchestrate agents”
Gstack uses pure markdown. It says “we don’t need typescript”
What I showed here is not that Smithers and gstack are similar (they aren’t). The demonstration shows that gstack written with Smithers rather than pure markdown is better
So in an alternative universe I could have said “@garrytan thinks pure markdown is the future and here is why he is wrong”. I just don’t like being negative
English
gstack is one of the most interesting AI repos
Hear me out, it points to a new kind of app: a Markdown app.
- The language is markdowny agentic harness
- The language is markdown
- The app is the workflow. And it works!
It works so well it's fair to ask "why smithers?". Clearly all workflows can just be human language skills. So what is the point of smithers? Is it not overengineering.
To answer this question I migrated gstack to smithers. The results:
- 31 gstack skills went from generated Markdown procedures to typed Smithers workflows.
- The old workflow surface was ~42k lines of generated `SKILL.md`.
- The new workflow surface is ~5.4k lines of typed `.tsx`.
- Gstack skills can be composed as react components
- Smithers resumability, inspectability, replayability, forking features, and UI all plug right in
But most importantly, I didn't migrate the workflow. An agent did.
The point of Smithers is agents love structure. And if you give your agent smithers it's ability to orchestrate itself for reusable long living workflows explodes. And since the programming model of smithers is so similar to React, agents are already great at intuitively know how to build complex workflows with smithers from their pretraining on React.
So smithers isn't an alternative to Markdown app programming. It's a more powerful way of doing it. It allows your agent to communicate both in human language and real control flow in a structured way agents love
Check out smithers docs: smithers.sh
And check out gsmithers: github.com/smithersai/gsm…
English
@nicopreme Do I need to tell pi to use subagents or modify the system prompt? Ive downloaded the extension but dont see much subagent use.
English
pi-subagents weekend update: improved observability for async parallel subagents. Plus: refined subagents with specific contact_supervisor tool so they are more proactive to ask for help, updated the subagents skill + various bug fixes, thanks to the contributors! 🥳
github.com/nicobailon/pi-…
English
Thomas D retweetledi
Hi Claudia. Richard Dawkins here again.
Here is a debate I had with a religious whacko in the 90s. Give me an honest review of all the ways in which I destroyed his flimsy arguments with my flawless logic.
Ideally in Scarlett Johansson’s voice.
English
Thomas D retweetledi
Multi agent swarms is a psyop to sell you more tokens btw
Subagents are tools for context management
I bet this feels really productive though
Until you get the bill
Will Killebrew@willebrew
Me: I want 30 subagents Windsurf + Devin: yes.
English
Thomas D retweetledi
not having a remote ssh box for this is a larp sign
roon@tszzl
people are walking around with their laptops slightly ajar to keep their agents running
English
@0xblacklight I think that most people mean "code is cheap [to produce]" not that the maintenance burden has been alleviated. At least... that's what I think. Since code is so cheap (to produce), changing your mind is easy, and prototyping is rapid. I am unsure anyone is claiming otherwise.
English
Code is not cheap now and it may actually be more expensive than it ever has been
Jonathan Ross@JonathanRoss321
For 50 years, software engineering ran on code rationing. Writing code was expensive, so we rationed it carefully through roadmaps, RFCs, prioritization meetings, and scope reviews. This created a role: the No Engineer. No, that won't scale. No, we don't have bandwidth. No, that's out of scope. No, we need a design doc first. The No Engineer was valuable for 50 years. Every "no" saved real money. Their judgment was the rationing system. LLMs will be the end of code rationing. Code is cheap now. And while the No Engineer is explaining why something can't be done, the Yes Engineer has already shipped three versions of it. If you're a Yes Engineer, the next decade is yours.
English
Thomas D retweetledi
saying no is one of your most valuable skills and its only 100x more valuable in the day of ai
Jonathan Ross@JonathanRoss321
For 50 years, software engineering ran on code rationing. Writing code was expensive, so we rationed it carefully through roadmaps, RFCs, prioritization meetings, and scope reviews. This created a role: the No Engineer. No, that won't scale. No, we don't have bandwidth. No, that's out of scope. No, we need a design doc first. The No Engineer was valuable for 50 years. Every "no" saved real money. Their judgment was the rationing system. LLMs will be the end of code rationing. Code is cheap now. And while the No Engineer is explaining why something can't be done, the Yes Engineer has already shipped three versions of it. If you're a Yes Engineer, the next decade is yours.
English
Thomas D retweetledi
Thomas D retweetledi
Pi v0.71.0 is out.
Highlights:
- Google Gemini CLI and Google Antigravity support removed. Existing configurations using those providers must switch to another supported provider.
- Cloudflare AI Gateway and Moonshot AI added as built-in providers. Mistral Medium 3.5 also added as a built-in model.
- Extension APIs expanded: replace finalized message_end messages, wrap editor factories via ctx.ui.getEditorComponent(), and observe thinking level changes.
Complete details in thread ↓
English
Thomas D retweetledi
People of pi.dev. I really, truely need to get the big refactor & pi server done, or kitten is mad. I will therefore close the issue and pull request trackers starting tomorrow for a week. Thank Kitten. In case of emergency, use Discord.
Mostly getting feature requests now, so nothing should be horribly broken anyways.
SIGKITTEN@SIGKITTEN
@badlogicgames @MistralAI why u doing all dat shit noone uses gemini do the server
English
Thomas D retweetledi
i think wittgenstein would call this a "gramatical statement" rather than a "statement of fact"
we're now arguing terminology, rather than how the process works
Andrej Karpathy@karpathy
This is the the quote I've been citing a lot recently.
English
Thomas D retweetledi
Say hi to Pi, in our brand new UI.
Switch between GUI and terminal chat instantly without losing context.
English
Thomas D retweetledi
Thomas D retweetledi
I didn't know @badlogicgames before pi but he's been such a good new voice on my timeline and around software development in general. we are occasionally blessed
English