118
Total Items Scanned
54
Fresh Candidates
15
Published
87.3%
Anti-Feed Protection
(Total - Published) / Total
| Source |
Input Share |
Output Share |
Delta |
| towardsdatascience.com |
5.6%
3
|
20.0%
3
|
+14.4%
|
| simonwillison.net |
1.9%
1
|
6.7%
1
|
+4.8%
|
| www.reddit.com |
92.6%
50
|
73.3%
11
|
-19.3%
|
Comparing the lowest-scoring articles that made the cut versus the highest-scoring ones that missed it.
✨ New Models & Weights
Last In (Lowest Score)
-
1605
Local AI: Managing VRAM by dynamically swapping models via API
-
1152
First AI implosion: Oracle
-
1046
how to train ai locally for creative writing
First Out (Highest Score)
-
965
2025 Open Models Year in Review
-
834
To Mistral and other lab employees: please test with community tools BEFORE releasing models
-
807
Baffled at stupidity of GPT-OSS-120b
🔬 Research Papers
Last In (Lowest Score)
-
1300
Ilya Sutskever is puzzled by the gap between AI benchmarks and the economic impact [D]
First Out (Highest Score)
-
639
[D] Causal ML, did a useful survey or textbook emerge?
-
629
[D] On the linear trap of autoregression
-
612
[P] Teaching AI to Beat Crash Bandicoot with Deep Reinforcement Learning
🛠️ Engineering & Code
Last In (Lowest Score)
-
2600
The Machine Learning “Advent Calendar” Day 14: Softmax Regression in Excel
-
2000
JustHTML is a fascinating example of vibe engineering in action
-
1030
Sick of uploading sensitive PDFs to ChatGPT? I built a fully offline "Second Brain" using Llama 3 + Python (No API keys needed)
First Out (Highest Score)
💬 The Discourse
Last In (Lowest Score)
-
606
One-Minute Daily AI News 12/14/2025
-
605
Built a pipeline for training HRM-sMOE LLMs
-
603
News Agent for Social Media
First Out (Highest Score)
-
54
Microsoft Scales Back AI Goals Because Almost Nobody Is Using Copilot
-
44
Simulated Company Shows Most AI Agents Flunk the Job
-
12
AI Agent Outperforms Human Hackers in Stanford Cybersecurity Experiment