ANYONE who claims you don’t need feature engineering doesn’t know what they are talking about.
@tunguz
-
The Holy Grail: Automated Table Combination for ML on Tabular Data
By
–
The whole field of structured (aka "tabular") data is still very poorly researched. One of the holy grails for the ML for tabular data is to have an automated system that can *combine* several tables, extract relevant information, and then create the best ML model for it. We are
-
Codex not recognizing CUDA on user’s machines
By
–
I still can’t get codex to recognize that CUDA exists on my machines.
-
Friday XGBoost models trained this week
By
–
It’s Friday. What XGBoost models have you trained this week? @trainxgb
-
NVIDIA ecosystem vs Mac or 5090 for large models
By
–
If you want ALL of the native support for the entire NVIDIA software ecosystem, AND you anticipate working with models that are 100 GB or larger, then it is a pretty solid choice. Otherwise a decently speced Mac would work, or a workstation with a 5090.
-
Codex 5.3 and Opus 4.6: optimal prompts guide
By
–
Codex 5.3 and Opus 4.6 have been released today. However, most people are using them wrong. Here is a quick thread to tell you what prompts you need to use to get the most out of them. A 1/69,420
-
True AI projects require iteration, not one-shot claims
By
–
My honest opinion: people who claim to have easily "one shotted" something with AI probably work on *very* conventional use cases. Or have ready made templates of steps that need to be accomplished. Every single project I've worked at with AI has resulted in lots of back and
-
Sonnet 5 dropping in 5 hours, v0 announces Build feature
By
–
BREAKING : SONNET 5 IS EXPECTED TO DROP IN 5 HOURS ACCORDING TO v0! v0 is about to drop their new “Build” feature too on top of Anthropic models. SOON x.com/v0/status/2019…
-
XGBoost endures while frameworks come and go
By
–
Frameworks come and go. XGBoost is forever. @trainxgb
-
Codex incorrectly suggests neural network for classical ML algorithm
By
–
I've been trying to reimplement a very simple *classical* ML algorithm (no, not XGBoost) in a different programming language. I've been using Codex for it. It keeps resorting to writing a *neural network* implementation, even though that approach is several orders of magnitude