The latest articles on DEV Community by T Mohamed Yaser (@yaser-123). https://dev.to/yaser-123 https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3382272%2F3f7a9ff2-d91b-47ef-8820-4d7c66b22bc1.jpg https://dev.to/yaser-123 en T Mohamed Yaser Sun, 05 Jul 2026 17:04:02 +0000 https://dev.to/yaser-123/we-stopped-passing-prompts-we-started-sharing-memory-3b3k https://dev.to/yaser-123/we-stopped-passing-prompts-we-started-sharing-memory-3b3k <h1> We Stopped Passing Prompts. We Started Sharing Memory. </h1> <p>Every multi-agent AI demo looks almost the same.</p> <p>One agent generates something.<br> Another agent consumes that output.<br> Another continues the chain.</p> <p>The longer the chain grows, the more context is lost.</p> <p>We asked ourselves one question:</p> <blockquote> <p>What if AI agents never passed prompts at all?</p> </blockquote> <p>Instead...</p> <p>What if they shared memory?</p> <h2> 🎥 Watch the Live Demo </h2> <p>See Syntapse in action as it transforms an idea into a complete software engineering workflow using shared memory powered by Cognee.</p> <p><a href="https://youtu.be/rS-OEcSBRpg" rel="noopener noreferrer">https://youtu.be/rS-OEcSBRpg</a></p> <h2> The Problem </h2> <p>Modern multi-agent systems suffer from what we call the <strong>Agent Telephone Game</strong>.</p> <p>Every agent summarizes information for the next one.</p> <p>Important architectural decisions disappear.</p> <p>Requirements drift.</p> <p>Small changes force the whole workflow to restart.</p> <p>For software engineering, this becomes a serious problem.</p> <p>A simple requirement like:</p> <blockquote> <p>"Replace PostgreSQL with MongoDB."</p> </blockquote> <p>can invalidate dozens of previous decisions.</p> <p>Instead of evolving...</p> <p>Most AI workflows regenerate everything.</p> <p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ft5eud2y1mi9mhzptu0oe.png" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ft5eud2y1mi9mhzptu0oe.png" alt=" " width="800" height="367"></a></p> <h2> Our Idea </h2> <p>Instead of prompt chaining, we built a <strong>Memory-First Operating System</strong> for software engineering.</p> <p>Every specialist reads from and writes to the same evolving memory.</p> <p>Nobody communicates directly.</p> <p>Memory becomes the source of truth.</p> <p>Our AI specialists are:</p> <ul> <li>💡 Ideation</li> <li>📄 Product Requirements</li> <li>🏗 Architecture</li> <li>⚙ Tech Stack</li> <li>🚀 Implementation Strategy</li> </ul> <p>Each specialist contributes knowledge to a shared graph powered by Cognee.</p> <h2> Memory Instead of Messages </h2> <p>Rather than:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code>Agent A → Agent B → Agent C </code></pre> </div> <p>We built:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> Shared Memory ↑ ↑ ↑ Ideation PRD Architecture ↑ ↑ ↑ Tech Stack Implementation </code></pre> </div> <p>Every specialist simply observes the latest state of memory and contributes when relevant.</p> <p>No prompt passing.</p> <p>No brittle orchestration.</p> <p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Foqpoepe8hashwofn8pbk.png" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Foqpoepe8hashwofn8pbk.png" alt=" " width="800" height="440"></a></p> <h2> Event-Driven Intelligence </h2> <p>Whenever memory changes...</p> <p>The system reacts automatically.</p> <p>If the user updates the technology stack:</p> <ul> <li>Memory changes</li> <li>Graph updates</li> <li>Relevant specialists wake up</li> <li>Implementation adapts</li> </ul> <p>Nobody presses "Run Again."</p> <p>Memory itself orchestrates the workflow.</p> <h2> Beyond Planning </h2> <p>Most planning tools stop after generating documentation.</p> <p>We wanted to go further.</p> <p>Once implementation prompts are generated, developers can use tools like Claude Code or Codex.</p> <p>After implementation is complete...</p> <p>The result is brought back into our platform.</p> <p>A verification stage checks whether the implementation still matches the original intent stored inside memory.</p> <p>If drift is detected...</p> <p>Instead of rebuilding everything...</p> <p>The platform generates:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code>Module 2 ↓ Module 2.1 (Repair) ↓ Updated Prompt ↓ Continue Development </code></pre> </div> <p>This allows projects to evolve instead of restarting from scratch.</p> <p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fbatttutzxonwrf7xxa22.png" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fbatttutzxonwrf7xxa22.png" alt=" " width="800" height="370"></a></p> <h2> Visualizing Living Memory </h2> <p>One of our biggest goals was making AI memory visible.</p> <p>We built:</p> <ul> <li>A live knowledge graph</li> <li>A synchronized memory lifecycle timeline</li> </ul> <p>Every operation—</p> <ul> <li>Remember</li> <li>Recall</li> <li>Improve</li> <li>Forget</li> </ul> <p>appears visually as the project evolves.</p> <p>Instead of watching logs...</p> <p>You watch memory think.</p> <p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fx7x7d3a0aft9b4hnbipa.png" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fx7x7d3a0aft9b4hnbipa.png" alt=" " width="338" height="806"></a></p> <h2> Why Cognee? </h2> <p>Cognee gave us something much more powerful than storage.</p> <p>It gave our AI specialists a persistent, evolving memory.</p> <p>That completely changed how they collaborate.</p> <p>Memory isn't just data anymore.</p> <p>It's the operating system coordinating the entire engineering lifecycle.</p> <h2> What We Learned </h2> <p>We started this hackathon thinking we were building another AI planner.</p> <p>We finished realizing we had built something much more interesting.</p> <p>The innovation wasn't adding more agents.</p> <p>The innovation was removing prompt passing.</p> <p>When every specialist shares the same evolving memory...</p> <p>AI behaves much more like a real engineering team.</p> <p>Thanks for reading!</p> <p>If you're building AI systems, I'd love to hear your thoughts on shared memory vs prompt chaining.</p> ai llm opensource hackathon T Mohamed Yaser Wed, 23 Jul 2025 15:49:15 +0000 https://dev.to/yaser-123/nochillai-built-during-exams-celebrated-with-a-community-fc https://dev.to/yaser-123/nochillai-built-during-exams-celebrated-with-a-community-fc <p><em>This is a submission for the <a href="https://dev.to/challenges/wlh">World's Largest Hackathon Writing Challenge</a>: Beyond the Code.</em></p> <p>Some projects are built with planning. Others are built with panic, pressure, and memes. NoChill.AI? Definitely the second one.</p> <h2> The Human Side of NoChill.AI 👨‍💻🧠 </h2> <p>I built NoChill.AI solo during WLH, right in the middle of semester exams. Between cram sessions and late‑night coding, I simply wanted to create something fun. The moment my chatbot fired its first roast—“Still up at 3 AM? Really?”—I knew I’d crafted more than code. I’d made an experience that shows up, even when real friends can’t.</p> <h2> A Meetup Moment I'll Never Forget 🎤✨ </h2> <p>I demoed NoChill.AI at a <strong>Code for India</strong> community meetup. Watching a room full of devs and students burst into laughter as the bot roasted volunteers was surreal. That instant feedback—real human laughter—made every all‑nighter worth it.</p> <h2> Bolt’s Discord: A Small Spark of Inspiration 💬 </h2> <p>While I didn’t dive deep into Bolt’s feature set, the <strong>Bolt Discord</strong> kept me motivated. Seeing others’ creative builds reminded me that even wild ideas belong in the hackathon.</p> <h2> What This Hackathon Taught Me 💡 </h2> <ul> <li> <strong>Collaboration with yourself:</strong> Balancing exams and solo coding taught me to prioritize and adapt on the fly. </li> <li> <strong>Fun has value:</strong> Not every AI needs to be serious—sometimes amusement is its own impact. </li> <li> <strong>Personal payoff:</strong> Building for myself—rather than a client—rekindled my love for pure creativity.</li> </ul> <h2> Shout‑Outs 🙌 </h2> <ul> <li> <strong>Code for India</strong> for the welcoming demo stage </li> <li>The <strong>Bolt Discord</strong> community for inspiration </li> <li>Everyone who volunteered to get roasted (and survived)</li> </ul> <p>NoChill.AI may never be VC‑backed, but it delivered real‑world laughs, genuine community moments, and the simple joy of building something just for fun.</p> <p><strong>That’s NoChill.AI: Built under pressure. Launched with laughter.</strong></p> devchallenge wlhchallenge community networking