A lot of people are arguing that HTML burns more tokens than markdown. It's true but you can save at least 40% by externalizing the CSS to a template with <link rel="stylesheet" href="./styles.css">. This style.css is your formatting so the LLM will never output CSS again.
I tested on a 12116 token HTML article and it dropped to 6,723 tokens so -44%!
You'll have this:
<div class="card">
<span class="badge">External CSS</span>
<h1>Hello, world.</h1>
<p>...</p>
</div>
Instead of this:
<style>
.card { /* 20 lines */ }
.badge { /* 12 lines */ }
h1 { /* 8 lines */ }
/* ...and ~100 more */
</style>
<div class="card">...</div>
@thekitze I think I will just copy and paste the node_modules from my previous pre-AI projects from here on out. But for now min-release-age = 1 month is all I am doing
@aschmelyun A lot of skills and MCP servers are gimmicks and not really practical gor coding. I use coding assistants on the daily and the only thing J have is Figma MCP which only helps in UI. You are doing fine.
If there is a repeated pattern of failure only then try exploring skills.
@dillon_mulroy We started with MCP servers to give external plug-n-play functionalities which had many flaws and frictions. SKILLS may not be perfect but they are definitely an upgrade.
The context overhead is actually far less almost negligible as compared to MCP
@nickstewart_@edzitron For people to use openclaw for something productive, they need such tasks to begin with. Everyone thinks they will become Tony Stark with it, nobody has Tony Stark level problems
@chribjel And also maybe a ledger filled with establishments and their numbers so that we can even find them easily.
Untapped market tbh, applying to YC rn as we speak
@chribjel Try coding detox. Leetcode or personal project without any vibe coding. Vibe coding follows similar patterns like any addiction. Mindfulness is key