working on GPT5 writeup for @latentspacepod
. by far I know the most common question is going to be "how does this thing do on tasks for larger, nonswebench codebases 1) compared to {competitor models} 2) foreach {competitor harness}?" willing to spend some time answering this
RESEARCH
-

GPT5 Performance Analysis on Large Codebases and Benchmarks
By
–
-
OpenAI Researcher Predicts Short AGI Development Timelines
By
–
oh shit openai researcher has short timelines
-
Lead Author Discusses Research Paper on X Platform
By
–
I didn't realize the lead author was also on X, here is his thread on the paper (and links): https://
x.com/mpshanahan/sta
tus/1950587921666281595
… -
8B Model Multi-Turn Tool Calling Capabilities
By
–
I'd really love to see an 8B model with reliable multi-turn tool calling, if one exists I've not experienced it yet
-
AI Alignment: Ontological Assumptions Shape Chatbot Worldviews
By
–
A new study argues that AI alignment is not just about human values, but a deeper exploration of ontological assumptions built into system architectures. Here, scholars make the case for considering ontology in AI design: https://
hai.stanford.edu/news/when-ai-i
magines-a-tree-how-your-chatbots-worldview-shapes-your-thinking
… -

Improving GenAI Model Evaluation: Rubric-Based Methods for Agentic Systems
By
–
Your GenAI model might be underperforming, and you wouldn’t know. Outdated eval methods miss hidden failure modes in agentic systems. Rubric-based evaluation fixes that. What it is & why it matters https://
snorkel.ai/blog/data-qual
ity-and-rubrics-how-to-build-trust-in-your-models/
… -

Chinese AI Labs Release Major Models: Kimi, GLM, Qwen
By
–
July has been a truly incredible month for model releases from China – Moonshot (Kimi K2), http://
Z.ai (GLM-4.5) and 5 new releases from Qwen I think it's undeniable that the best available open weight models now come from the Chinese AI labs -

Meta Developing Personal Superintelligence
By
–

BREAKING : Meta is building Personal Superintelligence! “Over the last few months we have begun to see glimpses of our AI systems improving themselves. The improvement is slow for now, but undeniable. Developing superintelligence is now in sight.”
-

MIT Research: How Language Models Track Dynamic Scenarios
By
–
How do language models track dynamic scenarios, like completing code or guessing your responses? MIT research finds that LMs don't track "state changes" step by step — they use mathematical shortcuts that we can control to boost LMs' prediction skills: https://
bit.ly/4lJfYfK -
Pelican Benchmark: Five-Attempt Model Selection Process
By
–
I've been meaning to do a version of the pelican benchmark where each model gets five attempts and then something picks the "best" one which moves on to the next round