So many of the less than ideal GSD quirks are because we are building for tools not built by us (Claude Code, Gemini CLI, Open Code etc). This is why we are building our flagship GSD CLI/TUI that is built from the ground up in Rust to provide the most seamless, zen-like and

The coolest thing about the GSD TUI I'm building is that it works by simply connecting to and communicating with the gsdaemon – a state-driven, AI development agent that runs 24/7 on your local machine designed from the ground up exclusively to build things following the GSD

the best passive income in crypto isn't staking it's owning 1% of every trade your community makes

We have decided GSD needs its own programming language for workflows. This will massively cut token costs as well as make it far easier to create custom "journeys" and "context packets" when using the GSD native TUI. We are creating the .arc filetype - which will serve as a

Every serious builder who has used Claude Code, Cursor, or any AI coding tool for more than a few weeks hits the same wall. The model is brilliant in flashes and incoherent across sessions. It can execute a complex task perfectly one day and go completely off-rails the next.

Open source is the best

You might consider creating a bespoke programming language tailored to your specific use case. Perhaps you might benefit from actually creating your own approach to storage with a custom database format. Those that truly get ahead in the AI coding landscape will be those that

x.com/i/article/2024…

With the latest news that Anthropic Max plans are only allowed to be used inside of Claude/Claude Code it's now becoming clearer than ever that GSD must evolve outside of the narrow ecosystem of proprietary coding model harnesses. We're on the right path to freedom, baby.

GSD has evolved so much since Lex Christopherson first posted it on Github. For this reason, we are redefining our mission as pioneering the agentic operating system space for reliable AI software development. By the end of 2026, GSD will be the de facto way that users build

The main question you should be asking is not "can the model do this?" but "should the model do this?" 9 times out of 10 when you're building agentic systems, the answer is no. You can probably just script it and it'll be much more reliable.

A lot of inspiration for the evolution of GSD comes from this motherfunker by the way 👏