Blog
Writing on GPUs, LLMs, MLOps, Kubernetes — and mindset · 3517 posts
#2026-03 765#english 592#culture 264#deep-dive 254#kubernetes 247#career 229#ai 216#llm 208#devops 193#2026-04 146#security 141#database 114#observability 113#communication 109#history 107#architecture 100#productivity 96#finance 88#economy 84#mindset 81#psychology 80#ai-papers 79#food 78#it 78#travel 78#deep-learning 77#japanese 77#networking 77#performance 72#business-travel 70#linux 70#gpu 69#ai-agent 66#cs-fundamentals 63#postgresql 60#rag 58#self-improvement 55#learning 53#mlops 53#ai-platform 51
The RISC-V RVA23 Profile — Distros Raised the Baseline, the Hardware Has Not Arrived Yet
RVA23 is the RISC-V application-processor profile ratified on October 21, 2024, and it is the first to make the Vector extension (V) and the Hypervisor extension (H) — both optional under RVA22 — mandatory. This matters
2026-07-16 · 17 min read #risc-v#rva23#ubuntu#toolchain#cpu-architectureSwift 6.3's Official Android SDK — What Shipped, and What Was Deliberately Left Out
Swift 6.3, released March 24, 2026, includes the first official release of the Swift SDK for Android. After ten years of community forks and unofficial toolchains, you can now produce Android binaries with a cross-compil
2026-07-16 · 12 min read #swift#android-swift#cross-platform#kotlin-multiplatform#swift-package-managerReact Compiler Got Ported to Rust — What Merged, What Did Not, and That "10x" Number
On June 9, 2026, PR 36173 — a Rust port of React Compiler — merged into facebook/react main. It added more than 120,000 lines across 461 files, and the number "3x faster as a Babel plugin, about 10x faster in the transfo
2026-07-16 · 19 min read #react#rust#compiler#performance#frontendProtobuf Edition 2026: Three Tightened Defaults, and the Schema Size Limits protoc Finally Enforces
Edition 2026 landed on protobuf.dev on July 13, 2026 — but v36.0-rc1 had already been published four days earlier, on July 9. Edition 2026 adds no new syntax; instead it tightens three defaults — a ban on name collisions
2026-07-16 · 19 min read #grpc#protobuf#api-design#serializationPython 3.15's abi3t — The Problem a Stable ABI Solves for Free-Threaded Builds, and What It Costs
With PEP 779, free-threaded builds shed their experimental tag in 3.14, but one thing besides interpreter performance still stood between here and Phase III, where free-threading becomes the default build — distribution.
2026-07-16 · 13 min read #python#free-threaded#cpython#concurrency#pep-703Rust 1.97 Cut Off Pre-Volta GPUs — Inside the nvptx64 Baseline Bump
Rust 1.97 (July 9, 2026) raised the minimum requirements for the nvptx64-nvidia-cuda target to PTX ISA 7.0 (CUDA 11 driver or newer) and SM 7.0 (Volta or newer). Maxwell- and Pascal-generation GPUs, and CUDA 10 and older
2026-07-16 · 13 min read #rust#cuda#gpu#compiler#nvidiaWhy Post-Quantum Certificates Are Not Here Yet — ML-DSA Signature Size, the 10kB Wall, and Merkle Tree Certificates
In the post-quantum transition, key exchange (encryption) is already well underway, but certificates and signatures are still at the starting line. The reason isn't cryptography — it's size. The ML-DSA-44 signature defin
2026-07-16 · 21 min read #security#post-quantum-cryptography#tls#webpki#certificatesRedis 8.8's Array Type — The First New Core Data Structure Since Streams, and What Valkey Doesn't Have
In May 2026, Redis 8.8 and Valkey 9.1 shipped six days apart. Redis 8.8 adds Array, a new core data structure built directly by antirez — a sparse index-addressed container for data where the index itself carries meaning
2026-07-16 · 16 min read #redis#valkey#data-structures#performance#open-sourcePasskey Portability Only Half Arrived — CXF Became a Standard, CXP Is Still a 2024 Working Draft
FIDO has two credential-exchange specs. CXF, which defines what gets exchanged, became a Proposed Standard on August 14, 2025 and even got an errata on March 9, 2026 — but CXP, which defines how it gets exchanged, is sti
2026-07-16 · 16 min read #security#passkeys#webauthn#fido#standardsPD Disaggregation Doesn't Increase Throughput — What Prefill/Decode Disaggregation Actually Buys You
PD disaggregation — splitting prefill and decode across separate GPUs — is a design that landed in vLLM, SGLang, and TensorRT-LLM alike in 2026, but wherever you look, all you find are numbers like "2x to 7x." Yet the of
2026-07-16 · 16 min read #llm#ai#inference#kv-cache#vllmWhy QUIC Gets Slower on Fast Internet — Receiver-Side CPU and the Kernel QUIC That Hasn't Arrived
QUIC clearly won on handshake latency and loss recovery — Google's internet-scale measurements show an 8.0% (desktop) drop in search latency and an 18.0% drop in video rebuffer rate as evidence. But once the link gets fa
2026-07-16 · 20 min read #network#protocol#performance#linux-kernelNix 2.35 Copies Flake Sources to the Store More Lazily — a 6-Year-8-Month Issue, and the Path Upstream Chose Instead of Lazy Trees
The first highlight of Nix 2.35.0, tagged July 13, 2026, is "Sources are copied to the store more lazily." The issue this one line closed, 3121, was opened by Eelco Dolstra himself on October 7, 2019, and closed on June
2026-07-16 · 14 min read #nix#reproducible-builds#build-systems#developer-experience#open-sourceOTel's Kubernetes Attributes Are Stable Now — What to Do Before k8sattributes Flips Its Default
OpenTelemetry's Kubernetes semantic conventions were promoted to stable in semconv v1.42.0 on June 12, 2026. But the Collector's k8sattributes processor still defaults to the old schema (v0), so most people haven't notic
2026-07-16 · 12 min read #opentelemetry#observability#kubernetes#semantic-conventions#telemetry-pipelineNative Histograms Are Stable Now — So Why Haven't You Turned Them On Yet
Prometheus native histograms showed up as an experimental feature in v2.40 (November 2022), went stable in 3.8.0 (December 2025), and 3.13, released July 1, 2026, is the first LTS to ship the feature in stable state. The
2026-07-16 · 17 min read #prometheus#observability#native-histograms#metrics#monitoringAWS Lambda MicroVMs — Running Agent-Written Code Under VM Isolation, and the New Problem Snapshots Create
AWS launched Lambda MicroVMs on June 22, 2026. It sells Firecracker micro-VMs as a service, and the market it's aiming at is clear — running user- or AI-generated code, isolated per tenant. It moves the container's share
2026-07-16 · 17 min read #security#firecracker#microvm#aws-lambda#sandboxingMCP Drops Sessions — Reading the Stateless Core in the 2026-07-28 Revision
The next MCP specification revision, 2026-07-28, is the biggest change since launch. Its core move is stripping state out of the protocol layer — the initialize handshake and Mcp-Session-Id sessions disappear, and every
2026-07-16 · 16 min read #mcp#ai#protocol#agents#integrationpg_plan_advice in Postgres 19: The Compromise From a Project That Has Long Refused Hints
The PostgreSQL project has refused optimizer hints for a long time. But PostgreSQL 19 Beta 1, released on June 4, 2026, added two contrib modules built by Robert Haas: pgplanadvice and pgstashadvice. This isn't Oracle-st
2026-07-16 · 14 min read #postgresql#query-planner#query-optimization#database#performanceParquet 2.13.0 Adds nan_count and IEEE 754 Total Order — How Float Statistics Got Fixed After 3 Years and 3 Months
Parquet's min/max statistics are the mechanism that lets query engines skip data. But on floating-point columns, this mechanism has been subtly broken for over 10 years — NaN returns false in every comparison, so it gets
2026-07-16 · 17 min read #data-engineering#parquet#columnar-storage#apache-arrowDo You Need a New Learning Rate When You Change LoRA Rank? — The Two Regimes μA (2026) Splits, and the Limits of That Evidence
Two competing beliefs circulate in practice at the same time: that changing the rank in LoRA means you have to re-find the optimal learning rate, and that using 1/r scaling makes the learning rate independent of rank. Th
2026-07-16 · 18 min read #llm#lora#fine-tuning#peft#trainingKafka Diskless Topics (KIP-1150): The Cross-AZ Cost You Trade for Latency — and a Feature Not Yet Shipped
On March 2, 2026, the Apache Kafka community accepted KIP-1150, Diskless Topics — a proposal to make object storage, not broker disks, the source of truth for data, eliminating the cross-AZ replication traffic that is th
2026-07-16 · 16 min read #kafka#streaming#distributed-systems#cloud-native