Learning Sudoku by Doing Gradient Descent on a Linear Program
Posted2 months agoActive2 months ago
mxkopy.github.ioTechstory
calmpositive
Debate
10/100
Artificial IntelligenceLinear ProgrammingMachine Learning
Key topics
Artificial Intelligence
Linear Programming
Machine Learning
The post discusses training linear programs using cvxpylayers to learn Sudoku, and raises questions about the limitations of current AI models in discrete settings. The author suggests that LPs can help bridge the gap in counterfactual reasoning.
Snapshot generated from the HN discussion
Discussion Activity
Light discussionFirst comment
33m
Peak period
1
0-1h
Avg / period
1
Key moments
- 01Story posted
Oct 20, 2025 at 10:11 AM EDT
2 months ago
Step 01 - 02First comment
Oct 20, 2025 at 10:44 AM EDT
33m after posting
Step 02 - 03Peak activity
1 comments in 0-1h
Hottest window of the conversation
Step 03 - 04Latest activity
Oct 20, 2025 at 10:44 AM EDT
2 months ago
Step 04
Generating AI Summary...
Analyzing up to 500 comments to identify key contributors and discussion patterns
Discussion (1 comments)
Showing 1 comments
mxkopyAuthor
2 months ago
In this post I go over training LPs using cvxpylayers, but I think the more interesting discussion lies at the end, where I argue that current AI models lack the ability to reason counterfactually in discrete settings & that LPs can bridge this gap.
View full discussion on Hacker News
ID: 45644123Type: storyLast synced: 11/17/2025, 9:06:50 AM
Want the full context?
Jump to the original sources
Read the primary article or dive into the live Hacker News thread when you're ready.