This post is my Agentic Coding 2025 Reading List: the best community-written posts I found this year on using LLMs and coding agents to write and ship professional code.
I organized them into following 9 categories based on the themes I kept seeing all year:
If you want to catch up on the year quickly, or you’re new to agentic coding and want a solid starting point, this list should give you the main ideas, debates, and failure modes without the noise.
The main trend this year is that my role (and a lot of other developers’ roles) shifted from writing code to reviewing it. That shift caused an identity crisis for some of us, plus real concerns about technical debt, security risks, and whether the promised productivity gains are real. This reading list captures how the community navigated these challenges.
Curation principles:
I aimed for ~50 posts (one per week). Ended up with 42 after aggressive filtering.
I only included community posts. No official launches, funding news, or company drama.
I optimized for practical value: workflows, failure modes, and lessons you can apply immediately.
Each link has a short "why it matters" note so you can triage fast.
I highlighted must-reads in each section.
Theme 1: Practical Workflows & Playbooks
Actionable patterns and tactics for shipping with agents
Claude Code: Best practices for agentic coding (Apr 18): The canonical starting point for Claude Code users.
Field Notes from Shipping Real Code with Claude (Jun 7): Adopt the anchor comment pattern.
How I program with agents (Jun 8): Mental model for AI agents.
Building a Personal AI Factory (Jul 1): Fix Inputs, Not Outputs.
Embracing the parallel coding agent lifestyle (Oct 5): What tasks can you parallelize across multiple agents without becoming a review bottleneck.
How I use every Claude Code feature (Nov 2): Reference for Claude Code’s entire ecosystem.
Ask HN: How can I get better at using AI for programming? (Dec 13): Tips from Hacker News community.
Theme 2: Case Studies: Wins and Limits
Real projects showing what worked, what didn't, and why
Human coders are still better than LLMs (May 29): Fixing bug in Redis. You still need to come up with novel approaches, not LLMs.
Building a Mac app with Claude code (Jul 1): Native macOS almost 100% built by Claude Code but the author has been building software for the Mac since 2008.
6 weeks of Claude Code (Jul 30): Years of tech debt cleared in a month and a half.
Pairing with Claude Code to rebuild my startup's website (Sep 20): Non-engineer founder's perspective.
Vibing a non-trivial Ghostty feature (Oct 11): Real feature shipped in 8 hours for $16; shows when to let AI prototype vs when to take over.
Claude Code can debug low-level cryptography (Nov 1): Debugging showcase; Opus 4.1 finds bugs in post-quantum crypto implementation.
I failed to recreate the 1996 Space Jam Website with Claude (Dec 7): Sometimes the simplest tasks are the hardest.
I ported JustHTML from Python to JavaScript with Codex CLI and GPT-5.2 in 4.5 hours (Dec 15): Showcase of code generation with minimal supervision when the task is constrained by a robust, implementation-independent test suite.
Theme 3: The Developer's Evolving Role
From code writer to reviewer
The role of developer skills in agentic coding (Mar 25): Situations where you need to supervise and steer the coding agents.
Why LLM-Powered Programming is More Mech Suit Than Artificial Human (Apr 21): Critical skill of "wielding the knife" to scrap and rebuild entire solutions.
Vibe engineering (Oct 7): When seasoned professionals accelerate their work with LLMs while staying accountable for the software they produce.
Theme 4: Tool Internals & Architecture
How AI coding agents actually work under the hood
How Cursor (AI IDE) Works (Mar 16): Self explanatory.
The Magic of Claude Code (Sep 30): Commands that power Unix happen to be perfectly suited for use by LLMs.
Understanding Spec-Driven-Development: Kiro, spec-kit, and Tessl (Oct 15): Evaluates SDD tools.
You Should Write An Agent (Nov 6): You only think you understand how a bicycle works, until you learn to ride one.
What if you don't need MCP at all? (Nov 2): Minimal CLI tools executed via Bash beat MCP servers for token efficiency.
Writing a good Claude.md (Nov 25): Self explanatory.
Theme 5: Code Quality & Technical Debt
The hidden costs of AI-generated code
Vibe code is legacy code (Jul 30): If you have to maintain it.
AI doesn't lighten the burden of mastery (Aug 17): Avoid the false mastery trap.
AI was supposed to help juniors shine. Why does it mostly make seniors stronger? (Sep 21): Where AI is good and where it falls short in coding.
Comprehension debt: A ticking time bomb of LLM-generated code (Sep 30): The Time you'll eventually pay to understand the code you didn't write.
Your job is to deliver code you have proven to work (Dec 18): Don't shift the burden of verification to reviewers.
Subscribe to get agentic coding updates delivered to your inbox:
Theme 6: The Productivity Debate
Measuring real impact
My AI Skeptic Friends Are All Nuts (Jun 2): Arguments against AI is a fad. Writing style makes it a delightful read.
Writing Code Was Never the Bottleneck (Jun 30): Argues code reviews, testing, debugging, coordination and communication were, and still are the actual bottlenecks.
Measuring the impact of AI on experienced open-source developer productivity (Jul 10): METR study found AI tools increased task completion time by 19%, though developers perceived a 20-24% speedup.
Where's the shovelware? Why AI coding claims don't add up (Sep 4): If we are 10x faster, why aren't we seeing an exponential rise in software releases?
Theme 7: Security Risks
Attack surfaces and defensive strategies
LLMs + Coding Agents = Security Nightmare (Aug 18): Introduces RRT (Refrain Restrict Trap).
Living Dangerously with Claude (Oct 22): Coins "the lethal trifecta".
Theme 8: Cognitive Friction & Limitations
Where current AI hits walls
Why LLMs can't really build software (Aug 14): They cannot maintain clear, dual mental models of both the requirements and the code's actual behavior.
Two things LLM coding agents are still bad at (Oct 9): No true copy-paste. Can't ask good clarifying questions.
Theme 9: Big Picture: The Future of Software
Industry-level shifts and what's coming
Learn to code, ignore AI, then use AI to code even better (Mar 27): Counters the narrative that new developers should forgo learning fundamentals and rely solely on AI.
Software in the era of AI (Jun 19): Did you go rewatch Rain Man after this Andrej Karpathy keynote?
Nobody knows how to build with AI yet (Jul 30): We're all junior developers again. In a permanent sense.
If you're going to vibe code, why not do it in C? (Dec 7): If AI writes the code, why use Python? We should return to C and Assembly for maximum machine efficiency.
Prediction: AI will make formal verification go mainstream (Dec 8): So far, for industrial software, the expected cost of bugs is lower than the expected cost of using the proof techniques that would eliminate those bugs.
Whether you're catching up during the holidays or bookmarking for later, I hope this list helps you build better software with AI in 2026.