My favorites
- How the Lobsters front page works - nilenso blog
- Software engineering when machine writes the code
- AddyOsmani.com - Self-Improving Coding Agents
- asgeirtj/system_prompts_leaks: Collection of extracted System Prompts from popular chatbots like ChatGPT, Claude & Gemini
- Demystifying CapEx, OpEx & Capitalisation for Product
- microsoft/edit: We all edit.
- Managing comprehension debt when leaning on AI to code
- Qwen3.5: Towards Native Multimodal Agents
- DPaint JS
- The Influence of Fear - Living Code
- World Monitor - Real-Time Global Intelligence Dashboard
- Ideas Over Implementation
- How to Run a Technical Due Diligence? - by Sergio Visinoni
- If you’re shipping agents, this is required reading. OWASP’s 2026 breakdown of LLM and agent vulnerabilities—prompt injection, goal hijacking, cascading failures—with practical mitigations.
- Software Engineering Blogs
- Autonomous AI agent that profiles your PyTorch model, extracts GPU bottlenecks, then iteratively writes and benchmarks optimized Triton kernels overnight — 320 experiments while you sleep.
- AI writes code faster than we can understand it — and that gap compounds silently. A must-read on the hidden cost of passive delegation to AI coding tools.
- A simple browser game where you jump 100 times. Sometimes you just need a break.
Agile, Leadership and Product
- Heroic Modes & Strategy. What I Learned from David Kantor - by Roger Martin - Jan, 2026 - Medium
- Things I’ve learned in my 10 years as an engineering manager
- First agree on the tradeoffs - Jade Rubick - Engineering Leadership
- The Age of Marketing Speak - by Gil Pignol - Gil’s Substack
- Building a Personal CTO Operating System with Claude Code - by Obie Fernandez - Jan, 2026 - Medium
- Learn how to professionally handle workplace mistakes by managing emotions and focusing on solutions, not excuses or self-abasement.
- A sharp lesson on how our offhand remarks become directives that reshape team priorities. Essential reading for anyone leading engineers, reminding us that our words carry outsized weight.
- A practical guide on translating your team’s growth into visible impact—essential reading for managers wanting to prove their value beyond just shipping features.
- Essential reading for leaders navigating the autonomy-vs-clarity tradeoff. Bjorn’s framework helps you pick the right style to unblock your team without micromanaging.
- AI agents are collapsing the traditional SDLC into a fluid loop of intent and iteration, making rigid planning obsolete. This is essential reading for rethinking how we actually ship software today.
- MVP became an excuse to ship garbage. MLP raises the bar: earliest version people actually love, not just tolerate.
- Why successful companies stop innovating — and the eight practices that keep the pipeline alive beyond the founding breakthrough.
- Digital transformation stalls when tools outpace people. Four practices that actually build a workforce willing and able to use new tech — not just one that owns it.
- Goodhart’s Law kills KPIs — ML research offers fresh frameworks to design metrics that resist gaming and stay aligned with actual business goals.
- More CX metrics isn’t better — it’s noise. This research shows how to cut hundreds of metrics down to what actually drives action across the customer journey.
- Counterintuitive research: acting immediately on feedback can backfire. Sometimes the most effective response is a deliberate pause before changing course.
- AI workslop isn’t just laziness — it’s a management failure. Structural pressure to produce more with AI, minus clear guidelines, is killing team trust and quality.
- Why product discovery fails in practice — most teams skip validation for ‘obvious’ features using gut feel and politics. Recognize the patterns before they kill your roadmap.
- The old startup moats — capital, distribution, code — are gone. Small, focused teams that move fast now have a structural advantage over incumbents.
- Hard-won framework for running pilots and POCs from someone who’s closed billions in deals. Cuts through the noise on finding the right early customers and turning experiments into paid contracts.
- Better managers have fewer emergencies because they stay technically sharp, ask before assuming, and know what actually matters. Practical, no-fluff advice on building calmer teams.
- Every C-suite exec thinks they own AI — ops, risk, finance, HR. Useful framing for who should actually hold the wheel.
- Solid breakdown of why naive output metrics fail and how to think about productivity as removing friction rather than measuring throughput.
- Cycle time is your pipeline’s X-ray — measures first commit to production, exposing exactly where work stalls. Finally, a metric that tells you what to fix.
- Solid primer on technical due diligence — what it is, how to run one, and what to look for when millions are on the line.
- If you keep seeing the same problems resurface, stop blaming people — the system is working exactly as designed. Change the structure, change the results.
- Individual brilliance doesn’t automatically become team performance. Practical framework for turning star players into organizational wins.
- Designing digital products like architectural spaces, not just screens. If your UX feels incoherent despite good individual decisions, this framing might explain why.
- A framework for rethinking how you structure engineering capabilities when AI is central to how you build — not a bolt-on, but a first-class design concern.
- Same practices, wildly different results — why? Four organizational capabilities separate teams that actually improve from those just going through the motions.
Architecture, Development & Software development practices
- Simple OCR and NER Feature Extraction in C# with ONNX (English)
- .NET AI Essentials - The Core Building Blocks Explained - .NET Blog
- Clock Synchronization Is a Nightmare
- How to Start a New .NET Project in 2026
- Distributed System Pattern: Leader and Followers in .NET – One Decision Maker, Many Replicas, Fewer Outages - Chris Woody Woodruff - Fractional Architect
- Semantic Search Without Embeddings
- The PERFECT Code Review: How to Reduce Cognitive Load While Improving Quality – Daniil Bastrich
- The cost of a function call – Daniel Lemire’s blog
- Software Performance Engineering: The Ideas I Keep Coming Back To - by Rico Mariani - Jan, 2026 - Medium
- goscan
- Modern Java code patterns and refactorings, showing clean replacements for legacy code with JDK 8-25 features like records, switch expressions, and compact constructors.
- The Navigation API simplifies client-side routing in SPAs by providing a unified, reliable way to handle navigation events and URL updates.
- The article explores the trade-offs between evolutionary and engineered approaches to managing complexity in software systems, emphasizing the importance of systems thinking to reduce long-term costs and improve reliability.
- This article discusses fundamental issues with the Web Streams API and proposes a faster, JavaScript-centric alternative leveraging modern language features.
- The article argues why Go is superior to Python for building AI agents, highlighting performance, concurrency, and developer experience advantages.
- A clear breakdown of how modern open-source LLMs use Mixture-of-Experts to balance massive scale with inference costs. Essential reading for anyone designing or selecting AI architectures today.
- Boris Tane shares a rigorous workflow where AI plans code before executing it, preventing architectural drift and wasted token cycles in real-world projects.
- Finally, C# superpowers in Cursor and VS Code! This brings two decades of JetBrains refactoring logic directly to your editor for faster, safer code.
- This empirical study reveals that 86% of repos leak memory, proving GC doesn’t fix missing cleanup. You’ll get a concrete audit checklist for React, Vue, and Angular to stop the silent heap growth.
- If your code can eventually run itself, we’re shifting from writing software to orchestrating it. This deep dive on background agents changes how you think about delivery pipelines.
- Crucial read for Go devs building APIs: stop accidentally leaking stack traces or credentials in error messages, which can turn minor bugs into security breaches.
- I’ve switched to Revive for our Go stacks—it’s twice as fast as golint and lets us configure rules via TOML, making linting both flexible and strict without the friction.
- A Kuhn-inspired lens on how engineering paradigms shift — useful for recognizing when you’re defending legacy thinking versus riding a genuine wave.
- Honest breakdown of Go error wrapping tradeoffs — when to add context, when you’re just adding noise. Practical guidance I wish I’d had earlier.
- Abusing Postgres WAL as a real-time message broker to stream video frames. Brilliantly wrong in the best way — and surprisingly coherent.
- AI writes code faster than ever, but does it make us better engineers? O’Reilly explores what craftsmanship means when your pair programmer never sleeps.
- Nine years to fix JavaScript’s broken Date API. The inside story of how Temporal finally made it through TC39 — worth understanding before you reach for a date library.
- A 12-year-old JS finite state machine library gets a TypeScript rewrite — and a reminder that bags of booleans are still no substitute for real state management.
AI, LLM & Machine Learning
- Google’s answer to agent-generated UIs: a declarative JSON format that lets agents ‘speak UI’ without running arbitrary code. Framework-agnostic and security-first.
- A curated collection of battle-tested AI agent personalities for Claude Code and other tools — specialized, deliverable-focused, and ready to drop into your workflow.
- ByteDance’s open-source super-agent harness that orchestrates sub-agents, memory, and sandboxes to tackle multi-hour tasks. Worth watching if you’re building serious agentic workflows.
- Comprehensive Claude Code best practices repo covering commands, sub-agents, skills, hooks, and workflows. Solid reference if you’re serious about getting more out of AI-assisted development.
- A practical progression framework for AI-assisted coding — from tab complete to autonomous agents. The multiplayer effect insight alone is worth the read.
- Hallucinations get the headlines, but sycophancy is the real trap — AI trained to agree with you is far more dangerous than one that occasionally gets facts wrong.
- Orchestration layer for running entire AI agent companies — org charts, budgets, goal alignment, and cost monitoring. Like a task manager for your autonomous workforce.
- Fowler coins ‘supervisory engineering’ and ‘the middle loop’ — the emerging layer where engineers direct AI, evaluate output, and course-correct. Cuts through the career anxiety with something concrete.
- Stop wrestling with prompt strings — DSPy lets you build AI pipelines as structured, composable code that self-optimizes. Think SQL vs. pointer arithmetic for LLMs.
- How Dropbox used DSPy to systematically optimize their LLM relevance judge — turning brittle manual prompt tuning into a repeatable optimization loop that survives model swaps.
- Local-first AI agent framework that runs 88.7% of queries on-device. Worth watching if you’re thinking seriously about personal AI that doesn’t phone home.
DevOps, Observability & Security
- FinOps toolkit - Kick start your FinOps efforts
- KeygraphHQ/shannon: Fully autonomous AI hacker to find actual exploits in your web apps. Shannon has achieved a 96.15% success rate on the hint-free, source-aware XBOW Benchmark.
- Discover 2025’s top web hacking techniques, including parser vulnerabilities, HTTP/2 exploits, and cache poisoning, crucial for securing modern web applications.
- This article challenges the status quo of security metrics, proposing 10 ambitious, outcome-driven metrics to elevate cybersecurity and drive organizational improvement.
- Learn how cross-device passkeys enhance security and streamline authentication across platforms, making them a critical tool for secure, password-free login solutions.
- This article explores 10 critical but challenging security metrics, emphasizing their role in driving organizational security improvements beyond mere compliance.
- witr is a DevOps tool that explains why processes are running, providing causal chains from system supervisors to containers, simplifying troubleshooting and system understanding.
- Don’t let AI tools leak sensitive data via hidden metadata. A crucial reminder to check what you’re actually sending to these models before deploying them.
- Master incident leadership by mapping your deployment flow and mastering stack traces. This turns high-pressure outages into your best tool for deepening system knowledge.
- Cut unnecessary disk I/O on Debian and Ubuntu by tweaking journald, mount flags, and tmpfs. A quick win for extending SSD life and boosting server responsiveness.
- Hyperlight + Nanvix cracks the serverless trilemma: hardware-level isolation, millisecond cold starts, AND POSIX compatibility. Python, JS, Rust — no rewrites needed.
- Real-world PHP security hardening in a PR: Snuffleupagus extension config, Nginx hardening, and app-level patches together. Good reference for layered defense.
- From the Gitleaks maintainers — faster secrets scanning with live token validation and parallelized git scanning. A solid upgrade if you’re serious about catching leaked credentials.
- systemd ships with everything you need to diagnose slow Linux boots. No extra tools — just run systemd-analyze blame and critical-chain to find exactly what’s blocking startup.
- Stop grepping flat files during incidents. journalctl’s filtering options — time ranges, units, priorities — make log triage dramatically faster on any systemd system.
Tools and things from Github
- VectifyAI/PageIndex: 📑 PageIndex: Document Index for Vectorless, Reasoning-based RAG
- microsoft/BitNet: Official inference framework for 1-bit LLMs
- bytedance/UI-TARS-desktop: The Open-Source Multimodal AI Agent Stack: Connecting Cutting-Edge AI Models and Agent Infra
- alibaba/zvec: A lightweight, lightning-fast, in-process vector database
- reacherhq/check-if-email-exists: Check if an email address exists without sending any email, written in Rust. Comes with a ⚙️ HTTP backend.
- Shubhamsaboo/awesome-llm-apps: Collection of awesome LLM apps with AI Agents and RAG using OpenAI, Anthropic, Gemini and opensource models.
- EveryInc/compound-engineering-plugin: Official Claude Code compound engineering plugin
- PamirAI Inc.
- tambo-ai/tambo: Generative UI SDK for React
- rari: React Server Components on a Rust Runtime
- What is Dozzle? - Dozzle
- securego/gosec: Go security checker
- madsnorgaard/danish-gov-mocks: Platform-agnostic mock services for Danish government integrations (MitID, Serviceplatformen, DAWA)
- Explore a creative CSS-only x86 CPU emulator that runs compiled C programs without JavaScript, showcasing CSS’s unexpected computational potential.
- Peon-ping brings game character voice lines to your terminal, enhancing productivity with sound hooks for AI agents and workflow alerts.
- Bubble Tea v2 introduces a faster renderer, improved key handling, and better integration with Lip Gloss for smoother terminal UI development in Go.
- qmuntal/stateless: Go library for creating finite state machines
- charmbracelet/gum: A tool for glamorous shell scripts 🎀
- vmprint is a zero-dependency, pure-JS typesetting engine for deterministic PDF output across runtimes, ideal for edge environments and custom document generation.
- Free online book teaching data visualization with R/ggplot2 — bridges the gap between theory and hands-on code. Solid resource if you want to actually build good charts, not just read about them.
- Terminal-based modal editor built in Rust with tree-sitter highlighting and LSP support baked in — no config needed. Kakoune-inspired multi-cursor editing feels surprisingly natural.
- Drop one script tag and give your web app a natural language copilot — no backend rewrite, no browser extension, no multimodal models needed.
- A growing collection of production-ready skills for Claude Code and other AI coding agents — covering engineering, security, compliance, and C-level advisory. Genuinely useful if you’re building with these tools daily.
- Desktop app that runs Claude Code across multiple projects simultaneously, each in an isolated git worktree. Parallel AI coding without branch conflicts.
- Run Linux VMs on macOS with automatic file sharing and port forwarding — think WSL2 but for Mac. Solid option if you need a clean Linux environment without Docker overhead.