Shane Larson

See the fastest LLM serving engine, 100% open source!!! If you’re running multi-round QA, RAG, or any long-context workload, you know the pain: slow time-to-first-token and expensive bills. You should take a look LMCache, a next-gen LLM serving engine extension designed to slash TTFT (time to first token) by 3–10x and boost throughput. How? LMCache stores KV caches from any reusable text (not just prefixes) across GPU, CPU DRAM, and disk, then shares them across serving instances. That means fewer recomputations, faster responses, and huge GPU savings. Key highlights: 🔥 Seamless integration with vLLM v1 ⚡ High-performance CPU KV offloading 🔗 P2P KVCache sharing across servers 💾 Storage on CPU, Disk, or NIXL ✅ Supported in vLLM production stack, llm-d, and KServe 📦 One-liner install: pip install lmcache If you’re scaling LLM inference, you should check it! 👉 GitHub link in the comments. #LLM #AI #Inference #OpenSource #RAG #AgenticAI

379

27 Comments

You may see folks claiming that they can get "A.I." coding assistants to do exactly what they want by completely and precisely describing the software they require up-front, and hailing the return of Big Design Up-Front. Anyone who's spent more than a handful of hours using these tools will know that this simply isn't true.

103

116 Comments

Hot take (with love): your To-Do app won’t get you hired. I’ve built them too. They’re great for learning CRUD, routes, and state. But the market’s seen a thousand. Hiring teams skim for signal. What they want to see: • You solved a real problem (yours or someone else’s) • End-to-end thinking (API design, auth, validation, tests) • You shipped (deploy + README + screenshots + a 60-sec demo) • You made it visible (why you built it, what’s next) It doesn’t have to be flashy—just useful and yours. How I’m doing it: FinLume. Not perfect. Not fancy. Still needs alot of work, But it solves my real headache (bills/subscriptions) and forced me into FastAPI, SQLAlchemy, auth, and testing. That’s the point. Simple non-To-Do ideas you could ship: • Renovation expense tracker with monthly summary (been living this! 🧱) • Job/quote scheduler for trades (my gardening days inspired this) • Interview tracker with follow-up email templates • Subscription cancel-reminder + renewal timeline Bottom line: Different > generic. Visible > invisible. Post progress. Ask for feedback. Make noise without screaming. #JuniorDeveloper #BuildInPublic #PortfolioProjects #LearnInPublic #PythonDeveloper #FastAPI

114

79 Comments

I just saved a client $900,000 by pressing DELETE. (Not joking.) We removed 1,000+ hours of custom code their previous vendor built. Now their system works better. The scam → $25,000 monthly "maintenance" fees. For 3 years straight. Just to fix bugs THEY created. The client literally cried when I showed them the truth... For two years, they'd been held hostage by their own ERP system. Paying their vendor to fix problems the vendor created in the first place. We took over. Dropped their billable hours from 1000 to 100. Per month. Here's what we found buried in their system: • Custom code for existing features • Over-engineered solutions • 90+ unnecessary apps They didn't need any of it. The hard part about our business is that ERP partners get paid by the hour. More custom code = more billable hours. Forever. It's a racket. When I showed them what we could do, their CTO literally said: "We paid $900k for features that already existed?" Yes. Yes, you did. We migrated their entire operation with 50 hours of custom code and $0 monthly maintenance fees. Now everything works better. Your offshore partner wants billable hours. More code = more problems = more hours. We want your business to work. Big difference.

355

39 Comments

The biggest lie in tech hiring: "We hire the best developers." No, you don't. You hire the developers who interview the best. I've worked with brilliant developers who can build entire systems from scratch but freeze up in technical interviews. They know every design pattern. They've solved complex production issues. They've mentored entire teams. But they can't explain FizzBuzz under pressure. Meanwhile, less experienced developers who practiced mock interviews land the $150k+ offers. Here's what I discovered after helping 100+ developers: The interview is a completely different skill than actual development. And most experienced developers never learn it. They think their GitHub speaks for itself. They assume their experience is enough. They walk into interviews unprepared. Then they wonder why they're getting lowball offers or straight rejections. The solution isn't becoming a better developer. It's becoming a better interviewer. After working across different continents with developers at every level, I've seen the patterns. The ones who succeed treat interviews like any other skill: They practice. They prepare. They get feedback. They improve. Want my complete interview preparation framework that helped developers increase their offers by $50k-$80k? Comment "INTERVIEW" and I'll send you the free guide. Stop letting poor interview skills cap your career.

96

171 Comments

After a chat with a great contractor yesterday, we got on the topic of T-Shaped profiles in contractors. In today’s fast-paced tech landscape, the most impactful contractors aren’t just specialists, they’re T-shaped. A T-shaped JavaScript contractor combines deep expertise in core JS technologies (React, Node.js, TypeScript, etc.) with broad capabilities across adjacent areas like DevOps, UX, testing, and architecture. They don’t just write code, they solve problems holistically. 🔍 Why T-shaped contractors stand out: Adaptability: They can jump into unfamiliar territory and add value quickly. Collaboration: They speak the language of designers, product managers, and backend engineers. Efficiency: They reduce bottlenecks by bridging gaps across the stack. Strategic thinking: They understand how their work fits into the bigger picture. Whether you're scaling a product, modernising legacy systems, or building something new T-shaped contractors bring the agility and insight to move fast without breaking things. 💡 Looking to hire? Prioritise depth and breadth. The best contractors don’t just code, they connect dots. #JavaScript #Contractors #Hiring #TechTalent #TShapedSkills #Frontend #FullStack #React #NodeJS #TypeScript #AgileTeams

30

1 Comment

styled-components has entered maintenance mode. Its maintainer was refreshingly honest: "Don't use this for new projects." Respect. But what about the millions of components already in production? My colleague Cody recognized the problem immediately. He submitted a PR with React 18 optimizations back in July 2024. With maintenance mode, that PR wasn't going anywhere. So Cody did what great engineers do. He turned our lingering PR into a fork, optimized it further, then opened it up for everyone. Linear tested it: 40% faster initial renders. No big code changes needed. The kicker? styled-components never implemented React 18's useInsertionEffect - a hook specifically designed for CSS-in-JS performance. It shipped in March 2022. Cody PR'd it fixed it in a day. We're calling it styled-components-last-resort because that's what it is. Not a new maintenance path. Not a forever solution. Just a performance bridge while teams migrate properly. Sometimes open source isn't about building the future. Sometimes it's about helping each other manage the present. Massive credit to Cody Olsen for the implementation, Kenneth Skovhus at Linear for testing, and Evan Jacobs for maintaining styled-components for years. Technical deep dive: https://lnkd.in/dk4CfsP8 GitHub: https://lnkd.in/duzY-Vy6 What's your styled-components migration plan?

111

3 Comments