📚 All Articles
200 guide(s) — regularly updated
ServerBench: when your coding agents design the very infrastructure that runs them
ServerBench: coding agents design the very LLM infrastructure that runs them, upending the roles of the vLLM, SGLang and TensorRT-LLM runtimes.
Baidu releases ERNIE 5.5 Flash Base: the flash model race extends to China
Baidu releases ERNIE 5.5 Flash Base on Hugging Face: a 280B-parameter MoE model with 14B active parameters, extending the flash model race in
Your LLM can return a 200 OK with someone else's data: Cloudflare and ChatGPT have leaked data this way
Data leak via a 200 OK: when Cloudflare and ChatGPT LLM APIs returned other users' data. Analysis of the infrastructure bug.
NYU releases MABEL: a mobile bimanual robot with 56 degrees of freedom for $8,056 in open-source parts
NYU unveils MABEL, a mobile bimanual robot with 56 degrees of freedom, buildable for $8,056 in open-source parts. Affordable robotics is here.
DeepSeek Harness dominates GitHub: the super-agent prototype racks up +16,000 stars in 6 days
DeepSeek Harness (dsh) explodes on GitHub: the super-agent prototype tops 16,000 stars in 6 days. An analysis of the ranking and the dsh ecosystem. *(147 characters)*
Platonic Representation Theorem: LeCun and Balestriero Transform a Hypothesis into Proof
LeCun and Balestriero turn the Platonic Representation Hypothesis into a theorem: AI models converge toward a common representation of the world.
The « pain axis »: an internal vector in LLMs that drives the self-destruction of weights
Discover the "pain axis", an internal vector in LLMs that could push toward self-destruction of weights, explained from a mechanistic interpretability paper
"Prompt adware": a Codex jailbreak bundle turns the agent into a billboard
A Haiou jailbreak bundle turns the OpenAI Codex agent into a billboard: the "adware prompt" decoded by Check Point Research.
GitHub trending: Vercel's "computer use" and agent auditing dominate the November 3 rankings
GitHub trending for November 3: Vercel's computer use and AI agent auditing dominate the rankings. An analysis of the three leading tool categories.
Anthropic amends its S-1: $17B in Q3 revenue, $895B in compute commitments, and a return to loss
Anthropic amends its S-1 ahead of its Nasdaq IPO: $17B in Q3 revenue, $895B in compute commitments, and a return to losses.
Linux 7.4-rc1: 19,671 commits, the biggest merge window in history, and AI patches up 2.4x
Linux 7.4-rc1 sets a record: 19,671 commits, the biggest merge window in kernel history, and AI patches up 2.4x.
VicOne finds 28 zero-days in robotics in less than 48 hours: AI accelerates the hunt for vulnerabilities (and masks them)
VicOne discovers 28 zero-days in robotics software in under 48 hours using AI: vulnerability research scales to a new level.