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
Bevy Game Engine — A Hands-On Rust ECS Tour of Modern Game Development (Deep Dive, 2026)
Inheritance-heavy object models from Unity and Unreal ruled the industry for thirty years, but the modern game engine of 2026 looks elsewhere. Entity-Component-System separates data from behavior and lays out memory for
2026-05-14 · 21 min read #bevy#game-engine#rust#ecs#gamedevBrowser Engines 2026 — Chromium, Gecko, WebKit, Servo, LadyBird: Who Is Actually Rendering the Web
Three engines render almost all of the web. Chromium owns more than 80%, WebKit takes a slice via iOS, and Gecko carries the last non-Chromium flag in single digits. Then 2026 arrives — Servo revives under LFE, LadyBird
2026-05-14 · 22 min read #browser-engine#chromium#gecko#webkit#servoAuth Provider Shootout 2026 — Clerk vs WorkOS vs Auth0 vs Stytch vs Better Auth vs Kinde vs SuperTokens, plus the NextAuth and Lucia post-mortem
The 2026 SaaS auth market has split in two. Clerk, Stytch, and Kinde won the B2C developer-experience lane. WorkOS owns the B2B enterprise-readiness lane (SSO, SAML, SCIM as an API call). Meanwhile, in self-hosted land,
2026-05-14 · 24 min read #auth#clerk#workos#auth0#stytchAuthoring Reviewable Pull Requests: Small PRs, Good Descriptions, and Stacked Diffs
Review latency is controlled by the author. Why small PRs win, the 'one PR = one idea' rule, the PR description as your most important artifact, making the diff readable, stacked PRs, self-review, responding well to feed
2026-05-14 · 24 min read #pull-request#code-review#git#stacked-prs#collaborationBuild Your Own Claude Code Skill — A Hands-On Deep Dive Into SKILL.md, Invocation, and Distribution (2026, english)
Agent Skills were introduced by Anthropic in October 2025 and opened as a standard in December 2025. As of May 2026 they are the default way to extend Claude Code. The mechanism is small — a SKILL.md with YAML frontmatte
2026-05-14 · 23 min read #claude-code#skills#anthropic#ai-tooling#hands-onAPI Schema Landscape 2026 — JSON Schema, OpenAPI 3.1, AsyncAPI, GraphQL, gRPC, Smithy, TypeSpec in One Map
In 2026 the real unit of an API is not the endpoint — it is the schema. JSON Schema 2020-12 became the shared base for OpenAPI 3.1, AI tool calling and config validation; AsyncAPI 3.0 settled in as the OpenAPI of events;
2026-05-14 · 23 min read #api-schema#json-schema#openapi#asyncapi#graphqlAI Search Engines 2026 Head-to-Head — Perplexity · You.com · Phind · Exa · SearchGPT · Gemini AI Mode · Kagi · Tavily, and the Deep Research Category
Search in 2026 is no longer 'ten blue links'. A new layer has slid in between the query and the answer — Perplexity's Pro Search/Deep Research/Spaces and the Comet browser, You.com's shifted positioning, Phind's develope
2026-05-14 · 29 min read #ai-search#perplexity#exa#searchgpt#deep-researchAI Video Generation in 2026 — Sora 2, Veo 3, Runway Gen-4, Pika, Kling, Luma, Hailuo, LTX (a deep-dive comparison)
The third leg of generative media has finally arrived. After music (Suno, Udio) and images (Midjourney, SDXL, FLUX), 2024-2026 was the year of video. Sora 2 reached consumer hands, Veo 3 shipped native synchronized audio
2026-05-14 · 19 min read #ai-video#sora#veo#runway#pikaProject Scheduling in the AI Era: Rebuilding Estimation, Velocity, and Sprint Planning From the Rubble
How AI coding agents broke estimation, velocity, and sprint planning, and what tech leads and EMs need to rebuild in their place. Covers why story points lost their meaning, how the bottleneck moved from implementation t
2026-05-14 · 26 min read #project-management#estimation#sprint-planning#velocity#ai-eraAI Image Generation 2026 — Flux, Imagen 4, Midjourney v7, Ideogram, Recraft, Firefly, DALL-E, Stable Diffusion: An Honest Comparison
The image companion to the AI music and AI video posts. The two real shifts in 2025-2026: open-weight Flux took SD-XL's throne, and Midjourney v7 reset the bar for consumer aesthetic. Imagen 4 GA, Ideogram v3 typography,
2026-05-14 · 25 min read #ai-image#flux#imagen#midjourney#ideogramAI Desktop Apps in 2026 — A Snapshot of Granola, Cleft, Lex, Highlight, Raycast AI, Ollama, and the Quiet Rise of the Ambient-AI Category
Spring 2026: the center of gravity for everyday AI use has quietly moved from chatbot tabs to desktop apps. Granola for meeting notes, Cleft / Superwhisper / MacWhisper for local dictation, Lex for AI-first writing, High
2026-05-14 · 36 min read #ai-desktop#granola#cleft#lex-app#highlight-aiAI Coding Assistant Limitations 2026 — An Honest but Fair Take on What Doesn't Work
AI coding assistants are powerful. But the shape of that power is not uniform. Deep bug debugging, large-codebase architecture, subtle performance regressions, concurrency correctness, genuinely new tech, multi-repo chan
2026-05-14 · 21 min read #ai-coding#ai-limitations#honest-take#ai-engineering#claude-codeAnatomy of an AI Harness — The Scaffolding That Turns a Model Into an Agent (Loop, Tools, Context, and Building Your Own)
The model is the engine; the harness is the car. Put a different car around the same engine and the driving experience changes completely. In 2026, the real center of gravity in AI engineering isn't the model — it's the
2026-05-14 · 16 min read #ai-harness#ai-agent#llm#agent-loop#tool-executionAI Coding Workflow Best Practices in 2026 — CLAUDE.md, AGENTS.md, .cursorrules, Skills, Subagents, MCP Deep Dive
The standards AI coding tools are converging on in 2026 — CLAUDE.md, AGENTS.md, .cursorrules, .github/copilot-instructions.md, plus Claude Code Skills, subagents, hooks, MCP, and permission modes. What goes in each file,
2026-05-14 · 20 min read #ai-coding#claude-md#agents-md#cursorrules#claude-codeAI CLI Coding Tools Shootout 2026 — Claude Code, Junie, Cline, Aider, Codex CLI, Cursor CLI, Continue.dev
The real battlefield in 2026 isn't the IDE panel — it's the terminal. We dissect seven terminal-native CLI harnesses that share the same models but behave completely differently: Claude Code, JetBrains Junie CLI, Cline,
2026-05-14 · 24 min read #ai-coding#claude-code#junie#cline#aider3D Development for the Web in 2026 — Three.js, R3F, WebGPU, Gaussian Splatting (a deep-dive on the modern web 3D stack) (english)
The WebGL era is over. As of January 2026, WebGPU is Baseline in Chrome, Edge, Firefox, and Safari 26, and web 3D has turned a new page. Three.js r182 recommends WebGPURenderer by default, TSL unifies WGSL and GLSL into
2026-05-14 · 18 min read #3d#web-3d#threejs#react-three-fiber#webgpuThe Complete Guide to AI Development Automation — GitHub Integration, Ticket-Based Agentic Workflows, Copilot, Claude Code, Devin, Jules (2025)
2025 was the year AI moved from 'suggesting code' to 'owning the ticket.' Assign a GitHub Issue to an AI agent and it rolls forward on its own — branch creation, implementation, PR, passing CI. A 4-level maturity model,
2026-05-14 · 29 min read #ai-development#automation#github#copilot#claude-codeVoice AI in Practice — The Complete Guide: Real-Time STT/TTS, Speech LLMs, Turn-Taking, and Deepfake Defense (2025)
Why "AI without a screen" became the hottest product category of 2025. Real-time voice pipelines (VAD/STT/LLM/TTS), speech LLMs (GPT-4o realtime/Gemini Live/Moshi), turn-taking and interruption, emotion and prosody contr
2026-04-15 · 12 min read #voice-ai#stt#tts#gpt-4o-realtime#moshiObservability 2025 Complete Guide: OpenTelemetry, Grafana / Datadog / Honeycomb / SigNoz, SLO and Error Budget, LLM Observability (2025)
Season 5 Ep 8. Without observability there is no operations, and without operations there is no product. Unifying the three OpenTelemetry signals (metric, log, trace), the Grafana stack (Prometheus, Loki, Tempo, Mimir) v
2026-04-15 · 11 min read #observability#opentelemetry#grafana#datadog#honeycombThe Complete LLMOps Guide: Three-Axis Versioning of Model, Prompt, and Eval Set, Canary, Cost Control, and the Platform Team (2025)
Building an LLM product fast has become easy. Running it sustainably is the hard part. Three-axis versioning across model, prompt, and eval set; Shadow/Canary/Blue-Green deployment; cost control through tokens, caching,
2026-04-15 · 11 min read #llmops#mlops#devops#canary#cost-control