Neo Kim

If you're serious about AI engineering (in 2026), then learn these 13 concepts: 1 How Vector Database Works → newsletter.systemdesign.one/p/what-is-a-ve… 2 How RAG Works → newsletter.systemdesign.one/p/how-rag-works 3 Design Personal Chat Assistant → newsletter.systemdesign.one/p/ai-chat-assi… 4 LLM Concepts - A Deep Dive → newsletter.systemdesign.one/p/llm-concepts 5 How to Design an AI Agent → newsletter.systemdesign.one/p/how-do-ai-ag… 6 What is Reinforcement Learning → newsletter.systemdesign.one/p/what-is-rein… 7 LLM Evals 101 → newsletter.systemdesign.one/p/llm-evals 8 Context Engineering 101 → newsletter.systemdesign.one/p/what-is-cont… 9 AI Coding Workflow 101 → newsletter.systemdesign.one/p/ai-coding-wo… 10 Agentic Patterns, Simply Explained → newsletter.systemdesign.one/p/agentic-desi… 11 How AI Agents Work → newsletter.systemdesign.one/p/ai-agents-ex… 12 Multi-Agent Architectures, Clearly Explained → newsletter.systemdesign.one/p/multi-agent-… 13 How MCP Works → newsletter.systemdesign.one/p/how-mcp-works What else should make this list? === 👋 PS - Want my System Design Playbook (for Free)? Join my newsletter with 200K+ software engineers now: → newsletter.systemdesign.one/join === 💾 Save now & repost to help others learn AI engineering. 👤 Follow @systemdesignone + turn on notifications.

@FrancescoCiull4 @DThompsonDev ah, love this!

Are you struggling to keep pace with AI? If you are a developer looking to stay ahead and do well in your job, this new one by @DThompsonDev is highly recommended. Huge respect for whoever finds the time and energy to write a book. Link below.

@swapnakpanda i'm not an expert, but i do 16 hours fasting often ^

Never ever fall for some nutritionist who suggest you to go for fasting to reduce your weight. Last year, I started doing intermittent fasting. First with 8 hours. Then increased it to 12. Then 16. Then 24. You won't believe. After 4 months, I started feeling weak. I didn't know why. Some people asked me if I have developed diabetes. I did blood sugar test. And yes, my sugar was in borderline. For the next 2 months, my sugar level went up only making me even more weak. I met a doctor. He told me clearly why you went for fasting? That too for 24 hours? Instead of fasting, concentrate on workouts. 4 to 6 hours fasting is fine. If you are non vegetarian, never go fully veg just because some one is suggesting. Your body needs protein. Your body may not be prepared for these weird changes. 1. Eat sufficiently. 2. Drink at least 2 to 3 ltr of water. 3. Never take soda, soft drinks. 4. Eat egg, paneer regularly. 5. Add fibre content in your food. 6. Work out for at least 30 minutes every day.

@RitikaAgrawal08 yes, and use "git apply" to unstash specific ones

Git Tip ✨ You can stash your changes with a custom message using "git stash push -m <message>" command. This helps in identifying and retrieving specific stashes later.

@csaba_kissi i'd use AI for prototyping and shipping outside core featurs ^

The real gain from AI is not typing speed, it’s faster iteration cycles.

@RaulJuncoV dual-ID approach is the cleanest way to balance performance and safety?

“UUIDv7 for everything” is starting to sound like the new default. I’m not sold. UUIDv7 fixes a real database problem. UUIDv4 is random, so inserts land all over the B-tree. That means: - Page splits - Cache misses - Write amplification - Slower inserts as the table grows UUIDv7 helps because it is time-ordered. New rows mostly land near the end of the index. Great. But now your ID carries timestamp information. That does not mean attackers can easily guess every future ID. But it can reveal metadata: Rough creation time Relative ordering Activity windows Traffic spikes Business volume patterns That may be fine for internal systems. It may be a problem for public APIs, invoices, orders, user URLs, or security-sensitive resources. My rule: - Use BIGINT when one database owns ID generation. - Use UUIDv4 or NanoID when the ID is public and should stay opaque. - Use UUIDv7, ULID, or KSUID when ordering and distributed generation matter. And for many SaaS apps, the boring answer wins: internal_id BIGINT PRIMARY KEY public_id UUIDv4 UNIQUE One ID for database performance. One ID for public safety. UUIDv7 is a good tool. But “v7 for everything” turns a good default into lazy architecture. Better IDs are still design decisions.

@RaulJuncoV excellent, yes--very important to save time + tokens + resources

@systemdesignone Context engineering is the one I’d double-click on. A lot of AI quality won’t come from better prompts, but from giving the model the right context, in the right order, at the right time.

If you're serious about AI engineering (in 2026), then learn these 13 concepts: 1 How Vector Database Works → newsletter.systemdesign.one/p/what-is-a-ve… 2 How RAG Works → newsletter.systemdesign.one/p/how-rag-works 3 Design Personal Chat Assistant → newsletter.systemdesign.one/p/ai-chat-assi… 4 LLM Concepts - A Deep Dive → newsletter.systemdesign.one/p/llm-concepts 5 How to Design an AI Agent → newsletter.systemdesign.one/p/how-do-ai-ag… 6 What is Reinforcement Learning → newsletter.systemdesign.one/p/what-is-rein… 7 LLM Evals 101 → newsletter.systemdesign.one/p/llm-evals 8 Context Engineering 101 → newsletter.systemdesign.one/p/what-is-cont… 9 AI Coding Workflow 101 → newsletter.systemdesign.one/p/ai-coding-wo… 10 Agentic Patterns, Simply Explained → newsletter.systemdesign.one/p/agentic-desi… 11 How AI Agents Work → newsletter.systemdesign.one/p/ai-agents-ex… 12 Multi-Agent Architectures, Clearly Explained → newsletter.systemdesign.one/p/multi-agent-… 13 How MCP Works → newsletter.systemdesign.one/p/how-mcp-works What else should make this list? === 👋 PS - Want my System Design Playbook (for Free)? Join my newsletter with 200K+ software engineers now: → newsletter.systemdesign.one/join === 💾 Save now & repost to help others learn AI engineering. 👤 Follow @systemdesignone + turn on notifications.

@petarivanovv9 thanks, covering this in a future newsletter

I'd also add cost modeling per request, latency budgets across model + retrieval + tools, eval discipline that survives a model upgrade, dataset curation, prompt versioning, and regression detection. These are the extra skills that actually separate AI engineers who ship reliably from those who don't.

@Tech_p001 appreciate it.

@systemdesignone Pure value content

@e_opore thanks, Dhanian

@systemdesignone Good compilation,Neo!

@petarivanovv9 just use a monolith for most projects

Most Node.js projects don't fail at scale. They fail before they get there... Not because of bad code. Because of bad structure. I've seen teams rebuild entire backends after 6-12 months, simply because the architecture couldn't evolve. The common mistake? Jumping straight into microservices. Yes, microservices scale well. But they also multiply: • deployment overhead • debugging complexity • coordination cost Before you need any of that. A modular monolith gives you the upside without the chaos. One codebase. Clear boundaries. Isolated modules. You get: • better developer experience • simpler testing • easier refactorings • a clean path to microservices later If a module needs to break out, it already knows how. That's the part most people miss. Most teams don't need microservices. They need discipline.

@twtayaan wait, microsoft has a distro, right? Azure Linux?

What if Microsoft made Linux

@e_opore i want an agent to debug for me

Copy-Paste Dev vs Debugging Dev

@socialwithaayan Grammarly is a must have, i try to keep as few as possible

50 CHROME EXTENSIONS THAT ARE ACTUALLY USEFUL 1. uBlock Origin — block every ad 2. Ghostery — see who tracks you 3. Privacy Badger — auto block trackers 4. ClearURLs — remove tracking from links 5. HTTPS Everywhere — force secure browsing 6. OneTab — collapse all tabs instantly 7. Toby — organize tabs in groups 8. Momentum — focused new tab page 9. Workona — workspaces for tabs 10. Session Buddy — save tab sessions 11. Grammarly — fix writing anywhere 12. LanguageTool — grammar in 30+ languages 13. Mailtrack — email open tracking 14. Simplify Gmail — cleaner inbox 15. Boomerang — schedule emails smartly 16. ColorZilla — pick colors from pages 17. WhatFont — identify any font 18. Wappalyzer — see any site's tech 19. Pesticide — outline CSS elements 20. Window Resizer — test responsive design 21. GoFullPage — full page screenshots 22. Nimbus — annotate screenshots fast 23. Loom — record your screen fast 24. Awesome Screenshot — capture and edit 25. Screenity — open source recorder 26. Raindrop — bookmarks that work 27. Save to Pocket — read articles later 28. Speechify — read pages out loud 29. Mercury Reader — distraction free reading 30. Zotero — organize research sources 31. Dark Reader — dark mode every site 32. Bitwarden — free password manager 33. Video Speed Controller — speed up videos 34. Unhook — remove YouTube clutter 35. News Feed Eradicator — kill social feeds 36. SimilarWeb — check site traffic 37. Tampermonkey — run custom scripts 38. ChatGPT for Chrome — AI in every tab 39. Merlin AI — GPT on any website 40. Compose AI — autocomplete your writing 41. Wordtune — rewrite sentences with AI 42. DeepL — AI powered translations 43. Perplexity — AI search assistant 44. Monica AI — AI copilot for browsing 45. Sider AI — ChatGPT sidebar for pages 46. MaxAI — summarize any page with AI 47. Harpa AI — automate web tasks 48. Tactiq — AI meeting transcripts 49. tl;dv — record and summarize meetings 50. SciSpace — understand research papers Your browser can do 10x more than you think. Most people never install a single one.

@twtayaan too late for my toddler, she just turned 4

Apparently this is the ideal age to begin DevOps

12 rules that'll help you reduce latency by 12x: 1 Database index 2 Compress payload 3 Group requests 4 Use HTTP 2 to send requests in parallel 5 Use CDN to keep data closer to users 6 Reduce external dependencies 7 Load balancer to distribute traffic 8 Scale vertically 9 Cache to serve popular data 10 Connection pooling 11 Message queue 12 Use efficient data serialization format What else should make this list?

👋 PS - Want my System Design Playbook for FREE? Join my newsletter with 200K+ software engineers right now: newsletter.systemdesign.one/join

@_jaydeepkarale indexing & query optimization is the first defence

@systemdesignone Connection pooling and indexing are must know

@swapnakpanda How many participants for this survey? I'm surprised not to see faang on top of the list?

🧑‍💼 India’s Top Workplaces for 2026: ➊ Infosys ➋ Accenture ➌ Amazon ➍ JPMorgan Chase ➎ SAP ➏ IBM ➐ Fidelity Investments ➑ Alphabet ➒ EY ➓ Wells Fargo TCS, the so called govt job, onsite hub, is not in this list. HCL, whose CEO dances on stage, is not in this list. Interestingly, only one Indian company makes it to this list. What does it say?

@CultureExploreX On my weekend read list

Before a child learned law, theology, or politics, he learned who received the first piece of bread. That lesson could happen in a palace, a farmhouse, a monastery, a Chinese banquet hall, a Bedouin tent, a Romanian kitchen, or a medieval great hall lit by rushlights. The table trained people in hierarchy, restraint, generosity, and belonging through ordinary acts: passing bread, honoring guests, seating elders, and refusing to grab the best portion. Roman banquets, medieval halls, Chinese New Year dinners, Islamic hospitality, Japanese tea ceremony, and family kitchens all carried the same hidden lesson. Manners were never just about politeness; they were a daily discipline that taught appetite to make room for another human being. When meals become rushed, private, and weightless, we do not only lose etiquette, we lose one of civilization’s oldest schools. newsletter.thecultureexplorer.com