One pattern I find useful for working with LLMs is a nice long ramble session. Sometimes the LLM needs more bits to understand what you're trying to achieve, but you're too lazy to type them. In these cases I like to lean back, switch to /voice and just ramble for like 10 minutes, total mess, anything goes, full stream of consciousness. Sometimes I declare it up top, something like "switching to speech recognition sorry for any typos...". Sometimes I turn it into a small interview of a few turns. But I find that the LLMs are somehow very good at reconstructing long incoherent rambles and often their echo of your own tangle of thoughts comes out quite a bit cleaner than what you started with. The result is that you improve the mind meld and have to correct things less from that point on.
Most developers still treat LLMs like magic autocomplete.
But the top 1% are already treating them like true agents that need engineered knowledge.
Context Engineering will help you ship faster with AI while still cutting down cost in 2026!
Hey @X algorithm 👋
Looking to connect with people who love building cool stuff.
If you're into:
SaaS
AI / ML
Blockchain
Full-stack
DevOps
Mobile Development
Data Science
Building in public
Let's connect. Always happy to meet fellow builders, learn from each other.
MCP !!😘
You spin up a tiny server that offers Tools, live Resources, and smart Prompts.
The agent discovers what’s available and uses them naturally when needed.
No massive context dumps. No repeated instructions. modular capabilities any host can plug into
Agent Skills might be the simplest good idea in AI right now.
Write your instructions once in a plain text file. Any agent (Claude, Cursor, Codex, Gemini CLI) reads it when relevant, ignores it when not.
No prompt copy-pasting. No token bloat. Just documentation.