RT @vincent_hus: Looking for open-source contributors. Apache 2.0.
841
Devyani Chavan retweeted
🚨 Anthropic just showed a 27-minute workshop on how to actually do prompts for Claude.
Taught by the people who built it.
Free. No registration. No paywall.
I've seen $300 courses that don't cover what they teach in the first 8 minutes.
Watch it and bookmark it now.
33
229
1,319
292,805
hardly solved the first one, it seemed logical and easy but when it came to putting that logic down in code..... aghhhhhhhhhhh
92
people will spend days trying to squeeze 0.1% more performance out of a model, meanwhile a quick data quality check or a thoughtful feature could give 5%. it's the 90/10 rule of ML: impact mostly lives in the grunt work, not the fancy bits.
1
71
so many 'new' llms are just deepseek v3 derivatives or similar. feels like the core architecture problem is mostly solved, and we're just fiddling with normalization for stability.
sebastianraschka.com/llm-arc…
73
everyone wants a perfect model, but getting it into prod with decent latency usually means ditching half the fancy stuff.
41
academically, ml students are sharp on models. practically, watching them debug a broken data pipeline or provision infra for their 'simple' model makes it obvious the actual systems building gap is still huge. kinda wild.
64
this constant push for the absolute bleeding edge model is exhausting. most of us are still just trying to get basic, *reliable* inference at scale without the whole thing falling over. incremental stability over speculative leaps, every time.
79