Devtools must be open source

8 hours ago

In the past, engineers rarely wrote personal software due to high costs and low ROI; they used config files and plugins instead.
AI agents now make personalization easy by automating source code modification, maintenance, and upstream synchronization via prompts.
Example: integrating meat.dev into the Shelley agent with a single prompt added pre-processing of diffs, saving time and effort.
Personalized software no longer needs complex plugin systems; agents can directly modify source code for individual needs.
This applies to small teams too, reducing the need for highly configurable off-the-shelf products.
Open-source agents like Shelley and Pi allow personalization, while closed-source agents like Claude Code limit it.

OpenMarkdown is a local-first, feather-light, and fast Markdown editor built for users and their agents.
It uses a native Rust core, ensuring quick loading times and native performance.
The editor is AI-free; instead, it runs on the user's existing agent, providing direct integration.

Micro-SaaS is obsolete; the new model is 'Service with a Software,' where private, overfit tools enhance a service rather than being sold as standalone products.
AI makes it easy to build small tools, but the value now lies in context like client relationships, trust, and customized fit, not the software artifact itself.
Independent builders can standardize meta-workflows (skills, prompts, pipelines) to efficiently create unique systems per client, reversing the old productization ladder.

Using CLAUDE.md or AGENTS.md files improves agentic tools like Claude Code and Codex by providing project context.
Manually updating these files becomes tedious and hard to maintain as project complexity grows.
A trick is to use agent logs (JSONL files stored in ~/.claude/projects or ~/.codex/sessions) to identify common problems with CLAUDE.md.

AI's PC era is arriving, focusing on personal AI that is easy, free, and dependable, operating locally on devices.
Personal AI should not be a cloud service but a layer between users and destinations, using user data and models.
Current software clutters screens to capture attention, with apps prioritizing platform goals over user intent.

<antirez>6 days ago

Traditional open source software follows a fixed development cycle with stable and unstable branches, where bugs are fixed before a release.
AI coding changes both software development and usage, as users can now ask AI to modify software directly.
Code repositories should serve as templates or examples for solving problems, allowing users to adapt code for specific needs.

OpenMarkdown is a local-first, feather-light, and fast Markdown editor built for users and their agents.
It uses a native Rust core, ensuring quick loading times and native performance.
The editor is AI-free; instead, it runs on the user's existing agent, providing direct integration.

Micro-SaaS is obsolete; the new model is 'Service with a Software,' where private, overfit tools enhance a service rather than being sold as standalone products.
AI makes it easy to build small tools, but the value now lies in context like client relationships, trust, and customized fit, not the software artifact itself.
Independent builders can standardize meta-workflows (skills, prompts, pipelines) to efficiently create unique systems per client, reversing the old productization ladder.

Using CLAUDE.md or AGENTS.md files improves agentic tools like Claude Code and Codex by providing project context.
Manually updating these files becomes tedious and hard to maintain as project complexity grows.
A trick is to use agent logs (JSONL files stored in ~/.claude/projects or ~/.codex/sessions) to identify common problems with CLAUDE.md.

AI's PC era is arriving, focusing on personal AI that is easy, free, and dependable, operating locally on devices.
Personal AI should not be a cloud service but a layer between users and destinations, using user data and models.
Current software clutters screens to capture attention, with apps prioritizing platform goals over user intent.

<antirez>6 days ago

Traditional open source software follows a fixed development cycle with stable and unstable branches, where bugs are fixed before a release.
AI coding changes both software development and usage, as users can now ask AI to modify software directly.
Code repositories should serve as templates or examples for solving problems, allowing users to adapt code for specific needs.

Hasty Briefsbeta