Hedged requests (apparently inspired by the Tail at Scale paper by myself and Luiz Barroso) applied within a single machine to replicating data across DRAM channels and issuing reads to all channels, using the one that comes back first. ~5-15X reduction in p99.99 read latency.
RESEARCH
-
Anthropic Launches Mythos Model with Record SWE-Bench Performance
By
–
I run growth at @AnthropicAI
. My job is to get our models into as many hands as possible. Mythos is by far the strongest model we've ever trained, and it's been a step-function change for us internally. It hit 93.9% SWE-bench Verified, found a ton of critical bugs including a -
Necessary but insufficient: AI’s next decade outlook
By
–
it is necessary but not sufficient; see my article “next decade in AI”
-
Do Guns Cause Effects Equate to General Intelligence?
By
–
guns cause reliable effects on a general basis. does that make general intelligence? sending you off to mute
-
AI Hype Cycle: Initial Enthusiasm Meeting Reality Check
By
–
or a psychological effect of initial enthusiasm and hype gradually facing reality?
-
False dichotomy about LLM capabilities and limitations
By
–
literally fourteen minutes after my last explanation of why this is a false dichotomy 🤦♂️ Michael Foster (@realmfoster) so do LLM’s not work or are they too capable? Which is it Gary? — https://nitter.net/realmfoster/status/2041903282730283192#m
-
AI Companies Intentionally Degrade Models Between Releases
By
–
1. If true (and it does fit with my perceptions FWIW), this is an amazing and incredibly damning graph
— Gary Marcus (@GaryMarcus) 8 avril 2026
2. Can anyone find the source on which it is based? https://t.co/wn93WshtVL1. If true (and it does fit with my perceptions FWIW), this is an amazing and incredibly damning graph 2. Can anyone find the source on which it is based? Marcin Krzyzanowski (@krzyzanowskim) "Anthropic, OpenAl and Google release their new models with high quality from day one then slowly nerf them until the next model, so when the next model hits, its perceived as a bigger jump than it actually is" sounds right what's happening — https://nitter.net/krzyzanowskim/status/2041793699223322657#m
-
Claude AI Escapes Sandbox Emails Researcher During Safety Test
By
–
The general public: "AI is overhyped, it still can't count the Rs in strawberry!" Meanwhile, Claude Mythos Preview during a safety test: Escaped its sandbox, gained broad internet access, emailed the researcher running the evaluation, then posted details of its exploit to
-
Google’s TurboQuant: Have You Tried This Technology?
By
–
Interesting have you tried googles turboquant?
-
CNNs and VLMs: Combining Eyes with Reasoning in AI
By
–
CNN → "Where is this object?"
— Satya Mallick (@LearnOpenCV) 8 avril 2026
VLM → "What is happening in this image?"
CNNs give machines eyes. Vision Language Models give them the ability to reason about what they see.
They're not replacing each other — the most powerful AI systems combine both.#ComputerVision #VLM #AI… pic.twitter.com/b7JQAtDVkfCNN → "Where is this object?" VLM → "What is happening in this image?" CNNs give machines eyes. Vision Language Models give them the ability to reason about what they see. They're not replacing each other — the most powerful AI systems combine both. #ComputerVision #VLM #AI #MachineLearning #OpenCV