The latest articles on DEV Community by Omar Megawer (@omar_megawer_bb207ad4b583). https://dev.to/omar_megawer_bb207ad4b583 https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3889387%2F7ebe17a5-f8ed-4109-8097-41b06e4601c5.png https://dev.to/omar_megawer_bb207ad4b583 en Omar Megawer Mon, 20 Apr 2026 16:55:09 +0000 https://dev.to/omar_megawer_bb207ad4b583/project-shadows-portable-agnostic-agents-35e0 https://dev.to/omar_megawer_bb207ad4b583/project-shadows-portable-agnostic-agents-35e0 <p>Building a multi-agent system around a different bet than most frameworks. Most agents jump straight to execution when you give them a task. People don't. We filter first, check if we're even the right person, look at what came before, then start.</p> <p>So I've been building agents where identity is a real filter (not a label in a prompt), context is narrowed per phase of work, and an agent can reject a task and hand it back if it isn't theirs. Nine specialized ones that push back on each other on strategy work.</p> <p><a href="https://omarmegawer.substack.com/p/part-3-project-shadows?r=4b5w1p" rel="noopener noreferrer">Check it out!</a></p> <p>Curious if anyone here has tried this kind of architecture and where you ended up.</p> ai agents memory openai