Keith De Alwis

I just claimed my .agent domain and joined the .agent community! get yours now and help shape the future of autonomous agents #X1Q6GJVU" target="_blank" rel="nofollow noopener">agentcommunity.org/join#X1Q6GJVU @agentcommunity_

@NPoverud Same 🙄

Claude Code seems to be down, and surprisingly few people seem to complain on X, strange.

A small thank you to everyone using Claude: We’re doubling usage outside our peak hours for the next two weeks.

Two weeks in. Under 4 hours sleep a night. Haven't seen the inside of a gym. Eating like absolute crap. Up at 6am and straight into build mode. The unlocks coming out of Claude Code and new agentic stacks right now are genuinely insane. I can't stop. I know I'm not the only one doing this. It's probably a 6 month window to build generational upside. But at some point we actually need to sleep. 😅

@Braah0x @openfangg It’s early but the dev has been cooking for a while. It’s security and trust looks impressive @openfangg you were about 3 months ahead of me 👏🏽

@keith_de_A Openfang looks good

It was 2am in Dubai. I was watching missiles get intercepted over the city. Not metaphorical ones. Actual ones. Lights tracking across the sky, the thud of the intercepts somewhere in the distance. Figured I wasn't sleeping. So I opened my laptop to try to fix the agent memory problem most of us have been quietly struggling with. Day 1: AI agents are brilliant. Day 2: they have no idea what they did on day 1. That's the dirty secret of multi-agent systems. Each session is a blank slate. Every context rebuilt from scratch. You're not running a team. You're interviewing new staff every morning. I got tired of it. So I spent the night fixing it. Persistent memory. Four runtimes. One architecture. And a stack that finally behaves like it actually remembers what it's doing. Here's what I built, what broke, and how you can get it. 🧵

@grankin_d 100% bro - extra sauce when you turn meeting meetings into markdown and load them straight the vault. I built a linear issue ingestion on the back of that flow so all actions nicely caught and linked across agents

landed on the same pattern - git-backed markdown, agents read/write as they go, not at session end. the part that still bites: what feeds the vault? code diffs, tickets, docs are easy to pipe in. but the decisions made verbally in meetings just vanish. that's the context source nobody's ingesting yet.

Full breakdown with the architecture, the trust bridge, the permission model, and why I didn't start with semantic search + How to get access to both Repos. de-alwis.com/i-fixed-the-du…

The real shift: agents write to the vault mid-session, not just at the end. Bug fixed? Note written immediately. Task handed off? Session log committed before the context window moves on. No panic at the edge of the context window. No twenty-minute recap next session. Just continuity. ♟️

Full list of what's trending in Claude Code MCP servers: mcpcat.io/guides/best-mc…

6. Figma MCP (Dev Mode) Design-to-code using accessibility trees. Cuts handoff friction massively. Built into Figma Dev Mode — enable it in your Figma settings. Your move. ♟️

New to Claude Code ? Here are my six MCP servers that actually matter. Not the full list of 9,000. Just the ones I'd install on day one. 🧵