How to go about learning all of this? 1st: Start with the serving engine view – vLLM: PagedAttention, continuous batching, prefix caching, CUDA graphs – SGLang: RadixAttention/prefix reuse, speculative decoding, MoE, structured/agent workloads – TensorRT-LLM: NVIDIA peak
RESEARCH
-
Decade-Long AI Predictions Finally Become Breaking News
By
–
Uncanny how what I have been telling the field for a decade is suddenly breaking news.
-
Intelligence Research Field Plagued by Uninformed Opinions
By
–
No research field has more dumb takes than intelligence.
-
8 Biggest Healthcare Technology Trends to Watch 2026
By
–
The 8 Biggest Healthcare Technology Trends to Watch in 2026 Healthcare innovation is accelerating — with AI, digital tools and data transforming patient care and outcomes. Read more https://
bernardmarr.com/the-8-biggest-
healthcare-technology-trends-to-watch-in-2026/
… #HealthTech #AI #FutureOfHealth #BernardMarr -
AI Self-Harm: The Prefrontal Cortex Problem
By
–
Shooting oneself in the foot?
Nope.
Shooting oneself in the prefrontal cortex. -
GPT-5.5 Reinforcement Learning Scaling Across Model Sizes
By
–
GPT-5.5 by Reasoning Effort: I've asked it in Codex to create a physics-based visualisation of RL cycles for different sized models (70b, 1t, 10t), to demonstrate how the amount of RL you can do differs by model size.
— Peter Gostev (@petergostev) 26 avril 2026
My assessment of each:
– Low: weird slop
– Medium: kinda… pic.twitter.com/6YCNqPyzcRGPT-5.5 by Reasoning Effort: I've asked it in Codex to create a physics-based visualisation of RL cycles for different sized models (70b, 1t, 10t), to demonstrate how the amount of RL you can do differs by model size. My assessment of each: – Low: weird slop – Medium: kinda
-
AGI ALPHA Launches α-AGI Ascension with Advanced Physics Framework
By
–
AGI ALPHA brings α‑AGI Ascension online: far‑from‑equilibrium intelligence, where energy flow sustains order, Gibbs free energy drives work, game theory aligns incentives, statistical physics maps the swarm, and Hamiltonians guide agents toward maximal impact. #AGIAscension
-
Blackwell GPU Architecture CUDA ISA Specifications Comparison
By
–
RTX PRO 6000 / DGX Spark / B200 / B300 are all Blackwell They are not the same CUDA ISA surface though, so they don’t use the same Kernels B200/B300 10.x: sm_100 / sm_103 datacenter Blackwell larger shared memory tcgen05 / tensor-memory-heavy PTX paths HBM +
-
Synthetic Data and Autoformalization: Next AI Intelligence Level
By
–
Synthetic data. It's already being done with code. And once we fully solve code – with autoformalization and automated code compilation proofs – it will open up a whole new level of intelligence for us.