ICML should be renamed ICBL (International Conference on Band-Aids for LLMs).
LLMS
-
Latent Space Podcast: GPT-5, StarGate, AI Engineers & More
By
–
gdb ep @latentspacepod we asked all your burning questions on gpt5, gptoss, stargate, RL, $300m ai engineers and more! thank you @lindsmccallum for helping us capstone the biggest launch week of the year*! (*until devday )
-
AI Benchmarks Will Continue Beyond Humanity’s Last Exam
By
–
“Humanity’s Last Exam” is probably not going to be the last AI benchmark…
-
Basic LLM Chat Terminal Interface Development
By
–
There's an "llm chat" basic terminal chat interface, nothing fancier than that yet though
-
Pixel Reasoner: Open-Source VLM Reasoning Framework Breakthrough
By
–
🚀 New Breakthrough in Vision-Language Reasoning 🧠🖼️
— Dr. Debashis Dutta (@debashis_dutta) 12 août 2025
Originally shared by @WenhuChen — Pixel Reasoner is the first open-source framework enabling Vision-Language Models (VLMs) to “think in pixel space” through curiosity-driven reinforcement learning.
💡 The Challenge
Most… pic.twitter.com/IXCFVuNMeMNew Breakthrough in Vision-Language Reasoning Originally shared by @WenhuChen — Pixel Reasoner is the first open-source framework enabling Vision-Language Models (VLMs) to “think in pixel space” through curiosity-driven reinforcement learning. The Challenge Most
-
LLM Command-Line Tool Releases GPT-5 Support Template Features
By
–
New release of my LLM command-line tool and Python library for interacting with Large Language Models – includes support for the GPT-5 model family and improvements to how tool calling works – you can now save multiple tool configurations in a template!
-
Anthropic reasoning traces differ from OpenAI DeepSeek approaches
By
–
When you look at the reasoning traces of Anthropic models, it just reads to me very unlike the OpenAI / DeepSeek reasoning. Anthropic looks a lot more like planning rather than eg deliberating on the answer, backtracking etc – like OpenAI and DeepSeek do, feels different
-
Opus 4.1 costs 3x more than GPT-5 and Gemini 2.5
By
–
I ran a semi-scientific cost test of 3x top models: Opus 4.1 was 3x the cost of GPT-5 and Gemini 2.5 Pro on real tasks. I gave the same prompt to each model (with reasoning enabled) and looked at the cost, without normalising for tokens. The point was to capture real cost and
-
Audio-Driven AI Model Supports Multiple Languages
By
–
yes, it's an audio driven model, feel free to upload audios in different languages
-
Stop Speculating: Wait for AI Models to Deliver Results
By
–
The same people who a week ago were saying that GPT-5 would be AGI and would destroy all the benchmarks are now doing the exact same thing with Gemini 3. Fuck, how annoying—wait for the models to actually come out and stop speculating. It's summer, enjoy the beach and grab