←
Neo Kim21,223 posts
Neo Kim avatar

Neo Kim

@systemdesignone

I Teach You AI Engineering & System Design • 0.5M+ Audience

◷ Joined April 2023▧ 2,213 Photos and videos
Neo Kim avatar
Neo Kim@systemdesignone·2h

I saved 3 hours a week on my calendar with Viktor. AI employees come in 3 levels, choose wisely: >Level 1: handles tasks you've already figured out >Level 2: figures out how to complete tasks you assign >Level 3: identifies problems, proposes solutions & takes action after your approval I spend a ton of time writing, building projects & managing tasks. So I hired Viktor (my Level 3 AI employee) to audit my Calendar. He reviewed my meetings, recurring commitments & fragmented work sessions. Then he helped prioritize my tasks using Eisenhower Matrix: >Urgent + important: Do first >Important + not urgent: Schedule >Urgent + not important: Delegate >Not urgent + not important: Eliminate I reviewed his recommendations & approved the calendar changes. RESULT: 3 hours of working time recovered per week. If you want more time for meaningful work, hire @viktor_com for your team. Start free. #AIemployee #Productivity Paid Partnership

0:53
Neo Kim avatar
Neo Kim@systemdesignone·3h

SOFTWARE ENGINEERS ONLY Why are there zero new JavaScript frameworks in 2026?

Neo Kim avatar
Neo Kim@systemdesignone·3h

Tell me if this sounds familiar. >You build AI agents for your team. >You expect them to work like real teammates. >Then you spend hours configuring bots, integrations & permissions. >It isn’t your agents. It’s the setup. Switch Buddy lets you create AI teammates in mins.

Say hi to Switch Buddy 👋 Claude Code writes code. Codex reviews it. Another agent debugs. No copying context or managing handoffs. Switch AI brings them into shared rooms with your team. Now Buddy helps you set it all up in minutes. Open source. Your agents. Your choice.

0:20
Neo Kim avatar
Neo Kim@systemdesignone·3h

How multi-agent orchestration works (explained in 2 mins with a bug fixing workflow): 1. Event Trigger: A new Sentry error triggers an AI agent. 2. Bug Investigation: The agent investigates & reproduces the error. 3. Code Generation: Agent then creates a fix & opens a GitHub PR. 4. Agent Handoff: The agent hands PR to a specialized PR agent. 5. Code Review: PR agent review the changes & fix issues it identifies. 6. Conflict Resolution: PR agent resolves merge conflicts if needed. 7. Code Merge: PR agent merges the PR once approved. But these agents need an orchestrator... Think of an orchestrator as a manager that coordinates specialized AI agents & delegates work. Composal Sparks takes this a step further: >Shared Agents: you can share agents & assign tasks through Slack. >Always-On Execution: Each Spark has its own cloud computer & keeps working after you log off. >Standing Orders: Agents follow persistent instructions, respond to GitHub events, Sentry errors, Slack messages. >Agent Handoffs: Specialized agents pass work to one another to complete engineering workflows. Each Spark acts as an orchestrator, so you don't need a single central manager. For software engineers & engineering teams already using AI coding tools, the goal is to automate the follow-through from issue to merged code. Composal Sparks is building this. Join waitlist now: composal.com/sparks

Today we're launching Sparks: multiplayer agents that orchestrate your entire software factory. Each one gets its own cloud computer and runs on the Claude, Codex or Grok subscription you already pay for. And they coordinate amongst themselves: one catches a production error reproduced it hands it off to a dedicated reviewer and merges the fix.

0:42
Neo Kim avatar
Neo Kim@systemdesignone·3h

if you want to 10x your productivity, try this

Post attachment

Notetaker is live today on iPhone. Now, your meeting notes can go wherever work takes you, for free. Use Notetaker on desktop for virtual meetings and on iPhone when you’re meeting in person. Your notes stay together across devices, so you can pick up right where you left off.

1:19
Neo Kim avatar
Neo Kim@systemdesignone·4h

I am begging you, learn how RAG works this weekend. Your AI agent is only as reliable as the data it retrieves. PDFs contain tables, charts, images & complex layouts that traditional RAG pipelines struggle with. Better document understanding = better retrieval & reliable AI agents. You'll be glad you learned this before building your next AI agent.

We’ve raised a $7.5M seed round led by Felicis so AI can read documents as easily as it reads databases. Finance, law and healthcare run on documents, and not databases like any other industry. Soon AI agents will read and write more documents than humans, and they’ll need to query them the way software queries tables. We’re building the computable layer that makes this possible. To do this at scale, we've brought together ML researchers and engineers from Google DeepMind, Cognition, and quant trading firms. We use computer vision and dual-stream models to turn complex multimodal documents into formats LLMs and AI agents can understand and reason over. If your team wants to turn messy documents into reliable, AI-ready data, book a demo: unsiloed.ai/book-demo

2:13
Neo Kim avatar
Neo Kim@systemdesignone·Nov 17

One of the most badass open-source AI inference projects in the AI era. >NVIDIA, AMD, Google, Meta, Intel, Thinking Machines Lab, Cloudflare, Modal & RadixArk team up to support SGLang >SGLang v1.0 was announced to run AI models at scale >RadixArk partners with Intel to make SGLang faster on Intel chips >SGLang adds support for Amazon's AI chips >RadixArk launches Miles v0.2 for AI model training SGLang Summit 2026.

2,000+ people. Two days. The first SGLang Summit. Together, we announced SGLang v1.0, released Miles v0.2, and launched the SGLang Consortium with nine founding members: @nvidia, @AMD, @intel, @Google, @Meta, @thinkymachines, @Cloudflare, @modal, and @radixark. SGLang is now used by 332 organizations. At Meta, it powers production inference for products serving billions of users. Teams from Modal, Cloudflare, LinkedIn, and others shared their production lessons. Across 90+ technical talks and six hands-on workshops, the community explored what comes next for inference, reinforcement learning, and AI agents. To everyone building, deploying, and contributing: thank you for making this possible. Explore the announcements, engineering lessons, and moments from SGLang Summit 2026 in the link below.

Post attachment
Neo Kim avatar
Neo Kim@systemdesignone·Nov 17

System design interview: A user updates their location at 10:00 AM. Your database confirms the write in 20ms. At 10:01 AM, the user opens the app on another device & sees their old location. Both devices are online. All database servers are healthy. What happened?

Neo Kim avatar
Neo Kim@systemdesignone·Nov 17

If you want to become a world-class AI engineer (in 7 weeks), then learn these 21 concepts: 1 LLM ↳ newsletter.systemdesign.one/… 2 AI Agent ↳ newsletter.systemdesign.one/… 3 RAG ↳ newsletter.systemdesign.one/… 4 Prompt Engineering ↳ newsletter.systemdesign.one/… 5 MCP ↳ newsletter.systemdesign.one/… 6 AI Coding Workflow ↳ newsletter.systemdesign.one/… 7 Fine Tuning ↳ newsletter.systemdesign.one/… 8 Vector Database ↳ newsletter.systemdesign.one/… 9 Multi-Agent Architectures ↳ newsletter.systemdesign.one/… 10 Context Engineering ↳ newsletter.systemdesign.one/… 11 Reinforcement Learning ↳ newsletter.systemdesign.one/… 12 LLM Inference ↳ newsletter.systemdesign.one/… 13 AI Agent Memory ↳ newsletter.systemdesign.one/… 14 Small LLM ↳ newsletter.systemdesign.one/… 15 AI Guardrails ↳ newsletter.systemdesign.one/… 16 Agentic Patterns ↳ newsletter.systemdesign.one/… 17 LLM Evals ↳ newsletter.systemdesign.one/… 18 A2A ↳ newsletter.systemdesign.one/… 19 AI Research Agent ↳ newsletter.systemdesign.one/… 20 Agentic Observability ↳ newsletter.systemdesign.one/… 21 LLM Judge ↳ newsletter.systemdesign.one/… === 👋 PS - Want my AI engineering playbook (for free)? Read right now: → systemdesign42.github.io/ai-… === What else should make this list? 1 Bookmark & RT to help others. 2 If you want to learn AI engineering, @systemdesignone will help you get there.

Post attachment
Neo Kim avatar
Neo Kim@systemdesignone·Nov 16

System Design Interview: Your database handles 50k requests per second. At 100k requests per second, latency jumps from 40ms to 3sec. So you partition the database across 10 servers. But 9 servers sit at 15% CPU. And one server hits 100% & rejects requests. What went wrong?