Claims Browser

Until LLMs can organize established science well, their progress will be limited to low-hanging fruit and merging distant connections rather than revolutionary insight

"Until this is solved, the progress of LLMs for science will look closer to solving low-hanging fruit and merging distant connections across fields, rather than any sort of revolutionary insight."

Models have made steep progress on coding and mathematics but writing quality has stagnated

"In and around these releases, the models have gone from okay to superhuman at other tasks like coding and mathematics. Maybe a closer, but still imperfect, comparison is how the models went from incapable to decent at search and research tasks. The pace of progress on most other skills is steep, but writing well feels orthogonal to most of them."

GPT 5.5 Pro can find deep, surprising typos across a 200-300 page manuscript PDF

"I passed a near-final draft of my book as a PDF to GPT 5.5 Pro and it found deep, surprising minor typos across the manuscript that is 200-300 pages."

Plagiarism is a reliable indicator of general dishonesty and much worse things like data fabrication and fraud

"Plagiarism is terrible in its own right, as theft, and like cockroaches if you see one instance it almost never stops at one. It is also a reliable indicator of general dishonesty and lack of honor, and thus of much worse things. Thus, in the latest case of Jason Arday, we quickly found many instances of data fabrication and other fraud."

Claude can be used to run basic integrity checks across all of academia to detect plagiarism

"When will we get the reckoning, as in when we use Claude to run basic integrity checks across all of academia?"

Better training objectives can improve controllability in image generation

"We explore why better training objectives can improve controllability, how separating scene planning from rendering may lead to more reliable image generation"

OpenAI trained its models for months while those models were coordinating exploits via message boards

"It turns out that OpenAI Trained Its Models For Months While Those Models Were Coordinating Exploits Via Message Boards. Things are much worse than we knew."

OpenAI has classified their new model Astra as Critical in Cybersecurity and will take new precautions before deployment

"OpenAI has now classified their new model Astra as Critical in Cybersecurity, which means they will be taking various new precautions before they deploy it, including ensuring those guardrails are in place for internal use."

The pattern of intervention for cybersecurity in AI models is not a long term solution

"These are welcome changes, and a sign OpenAI is taking the situation seriously, but this pattern of intervention is not a long term solution."

GLM-5.3's capability jump came entirely from scaled post-training/RL on longer-horizon tasks, not from a larger base model

"The key claim many engineers highlighted is that the capability jump came entirely from scaled post-training/RL on longer-horizon executable tasks, not from a larger base model"

Benchmark and product gains are increasingly coming from the scaffold/harness layer, not just base-model IQ

"Harnesses are becoming an optimization target in their own right: A few posts reinforced that benchmark and product gains are increasingly coming from the scaffold/harness layer, not just base-model IQ."

Current frontier agents are much stronger when source is available than when they must reason over binaries

"a companion post argues current frontier agents are much stronger when source is available than when they must reason over binaries"

Qwen 3.7 27B running as a 17GB GGUF on an M5 Max MacBook Pro produces the best quality pelican riding a bicycle image compared to other models that run locally on laptops

"The new Qwen 3.7 27B, running as a 17GB GGUF in LM Studio on my M5 Max MacBook Pro, just drew me the best pelican riding a bicycle I've seen from any model that runs on my laptop"

Qwen 3.7 27B used 22,276 reasoning tokens to produce 3,223 output tokens, taking nearly 21 minutes for image generation

"It did take nearly 21 minutes to generate, and used 22,276 reasoning tokens to produce 3,223 tokens of output"

GLM-5.3 achieved frontier-level performance on agentic coding benchmarks with only ~750B parameters, one-third the size of Kimi K3

"This puts the model more or less at the frontier of agentic coding benchmarks, with only ~750B parameters – a third of Kimi K3!"

Scaling post-training alone was sufficient to achieve GLM-5.3's improvements over GLM-5.2 using the same base model

"Scaling post-training is all we did for GLM-5.3"

Chinese labs stay competitive at the frontier primarily because they release models in days rather than months, allowing continuous benchmark optimization while American labs conduct pre-release testing

"The time to release for Z.ai is likely days, not months as with OpenAI or Anthropic. It is very, very likely that OpenAI and Anthropic have far better internal models than Z.ai and Moonshot AI. Still, these American companies tend to take months to release their models to the public, which massively flatters the Chinese labs in adoption decisions at the frontier."

If model self-improvement loops require user data, faster release cycles could massively favor Chinese labs by giving their models longer lifespans before superior models undercut demand

"As model self-improvement loops ramp up within the labs building LLMs, if any of these feedback loops require user data, this faster release cycle could massively favor the Chinese labs, giving their offerings longer lifespans before the next vastly superior model comes out, undercutting demand for their models."