Maciej Winnicki
@mthenw
thinking fast and slow (mostly slow)
Some days it is kind of depressing watching the idiocracy of modern tech influencers teach devs nonsense that probably comes from job-hopping and never seeing the consequences of actions. We know how to build better software; it's not what the popular idiocracy promotes though
Great talk for sure, but how are all those analogies actually useful? LLM=OS, cool, but what’s the practical takeaway?
Andrej Karpathy's (@karpathy) keynote yesterday at AI Startup School in San Francisco.
Looks cool! But my ultimate goal is to be the last @arcinternet user on Earth. Unless someone from @browsercompany shows up at my place and uninstalls it from my computer, I’m not doing it.
Move with me @diabrowser
Join us at Evals track at 11:45! We’ll be speaking about our failures and lessons from maintaining thousands of agentic evals and how you can avoid them
Is it just me, or did Cursor start randomly removing unrelated lines of code while editing a file?
TL;DR: We built a transformer-based payments foundation model. It works. For years, Stripe has been using machine learning models trained on discrete features (BIN, zip, payment method, etc.) to improve our products for users. And these feature-by-feature efforts have worked…
April was my last month at Vendr, so I'm on the lookout for a new role! LinkedIn in the reply 👇 but here is a quick summary: • 13 yrs in dev (TypeScript/React, Go, AWS, Cloudflare) • 5 yrs leading teams across the US & Europe — I know how to build high-performing teams and a…
One of the few AI/LLM use cases that actually works and I enjoy interacting with. If I see kapa on tech docs site, I have much more confidence I will find an answer.
This is your technical support on @kapa_ai. Our customer deployed in January, and they saw a ~30% reduction in support tickets. This isn’t from making support harder to get to, or impossible to log a ticket with, or anything like that. It’s simply adding an AI assistant that…
yup, I agree
"companies that don't hire remote are evil" "ok fine we'll hire remote, say hi to maciej your new polish teammate, he's smarter and less annoying than you" "wait no not like that"
Anyone here deployed LanceDB to AWS? Would love to talk with you.
Looking for my new ideal vector+text store for RAG because I can't use @lancedb in Cloudflare Workers. Must: - cosine similarity + BM25 - Typescript SDK - fully managed (no managing vms/machines) - unlimited namespaces/tenants/logical isolation - pay per use - can deploy in EU
Claude + RAG over your Apple Notes? No one asked but I did it. Now you can talk with your notes using Claude Desktop app thanks to "Model Context Protocol". Fully on device, completely private/offline, no managed services/API keys needed. Link below 👇
If you want to try out the new Workers Static Assets feature with Remix, check this out: github.com/remix-run/remi…