🤖 Generative Programmer | Issue #10

AI-Assisted Development, Prompting & Autonomous Agents

📣 Welcome back to Generative Programmer!

This week: the new AI model landscape, making LLMs write better code with a simple prompt, and the debate—one long context thread or many short chats Bookmark GenerativeProgrammer.com and share this issue: 🐦 Twitter, 💼 LinkedIn, 🦋BlueSky, ✉️ Email Let's dive in!

🚀 This Week's Top AI Picks For Devs

🔄 4 Patterns of AI-Native Development
👌 Agentic Workflows Patterns Explained Simply
📋 AI Native Developer Tools Landscape
⚠️ Can LLMs Write Better Code, Really?(yes, if you ask them)
🧩 Gemini 2.5 Pro Gets Stress Tested
👨‍💻 Leading Eng Teams using AI
🚨 AI Blindspots That Break Your Coding Agents
🎯 Why Agent Experience Is the Next UX Paradigm
💬 Prompt Engineering Secrets from Top AI Agents

…and 35 more updates to explore! 👇

🎯 Featured Article

What Are Agentic Workflows? A comprehensive primer on agentic workflows—what they are, how they work, and why they matter.

The post breaks down core agent components (reasoning, tools, memory), key patterns like planning and reflection, and real-world use cases including agentic RAG and coding assistants. It also compares agentic vs. non-agentic workflows and explores benefits, trade-offs, and architectural considerations.

🗞️ 📰 Long-Form Reads

Patrick Debois on AI Native Dev Four key patterns reshaping software roles in the AI era: from coding to managing agents, intent over implementation, discovery-first workflows, and knowledge as leverage.

Sid Bharath on Model Context Protocol
Part 1 of a 3-part series on MCP—how it enables secure, real-world actions for AI agents beyond simple API calls. Includes setup steps, use cases, and why it matters. 🧩

Can LLMs write better code if you keep asking them to 'write better code'? - Repeatedly asking an LLM to "write better code" might not yield meaningful improvements, as this vague instruction lacks specificity about what "better" means in your context. 🤖✨

Putting Gemini 2.5 Pro through its paces - Simon Willison test drives Google's new Gemini 2.5 Pro model, described as "a thinking model" for complex problems that's already topping the LM Arena leaderboard.

Leading Effective Engineering Teams in the Age of GenAI -

Addy Osmani

explains that engineering leaders should view AI as a junior team member requiring guidance rather than a tool for producing more code faster, emphasizing the importance of defining what "better software" means and adopting a "trust but verify" approach to AI-generated code.

Rewilding Software Engineering: Explainability through contextual tools - Simon Wardley explores how contextual tools can help align mental models with software system realities, featuring case studies that demonstrate how dynamic tools address complex problems and enhance decision-making. 🔍

20 AI Blindspots - The article mentions blindspots the author has observed in LLMs during AI coding, and indicates possible future suggestions for Cursor rules to address these issues. 🤖

Peter Yang on Creator Economy: A practical 2025 guide to picking the right AI model—Claude for writing, ChatGPT for voice/research, Gemini for long video, Grok for free power use, Perplexity for fast lookups.

Why Agent Experience Matters - As AI agents become autonomous actors in digital systems, developers should prioritize designing interfaces and APIs that are easily navigable by LLMs, not just humans, ensuring their software remains relevant and powerful in an agent-driven ecosystem. 🤖

Load Testing LLMs - When load testing LLMs in Kubernetes, consider using existing load testing tools with minor customizations rather than building from scratch, as Christian Posta found about 90% of their custom-built tool's functionality was already available in standard tools.

Prompt Engineering for Unbeatable AI Agents - The author observes that top-performing AI agents like GPT Pilot, GPT Engineer, Devin, and Self Operating Computer all employ similar prompting techniques to achieve their cutting-edge results. 🤖

The New Stack on Kagent Solo.io launches Kagent, an open-source AI agent framework for Kubernetes. Built on AutoGen, Kagent helps DevOps teams run AI agents for debugging, deployment, and security on K8s.

Accelerate Spring AI Development with Effortless Privacy from CodeGate - A quick demonstration of using CodeGate, an open source local AI Gateway, to protect personally identifiable information while working with Spring AI and various AI coding assistants. 🛡️

How to adopt Gemini Code Assist – and measure its impact - Google's blog post promotes their AI coding assistant with impressive adoption statistics, but conveniently skips over some of the concerns about code quality, security vulnerabilities, and developer deskilling that come with AI-generated code.

🐦 Social Highlights

Paul Copplestone on X Supabase saw 100% Vercel Marketplace signups after launching a v0 integration.

Simon Wardley on LinkedIn "Vibe coding" delivers fast dopamine, slow debt. A sharp take on AI-assisted dev, tech cycles, and retraining over replacement. Don’t fire your engineers for AI. Wardley warns it’ll backfire—again. Retrain instead.

Ruzgar Zere on LinkedIn A practical framework for decomposing PM workflows into AI agents — with clear steps, real examples, worth a skim.

Andrej Karpathy on X Long LLM threads as moats? Karpathy breaks down why infinite context isn’t always better—speed, attention, and training mismatch all play a role.

elvis (Omar Sar) on X LLMs respond to vague iteration prompts like “write better code” with real improvements—proof that prompt loops can drive quality gains.

AI Tools for Product Managers - While AI won't replace Product Managers, the post provides 19 specific AI tools across research, prototyping, analysis, and design that PMs can incorporate to enhance productivity and increase personal expenses.

12 rules to vibe code without frustration - Peter Yang shares a guide with 12 rules for coding with AI based on his 50+ hours of app building experience, noting that AI tools actually made him more technical as he learned to supervise AI 💻

Nikki Siapno on LinkedIn
A visual 10-step guide to building stateful AI agents with memory—covering models, workflows, and Zep integration.

🛠️ Tech & Tools Worth Exploring

Vector DB Comparison - The most comprehensive comparison of vector databases

LLM Leaderboard 2025 - Verified AI Rankings - A resource for comparing AI models across various metrics including benchmarks, pricing, capabilities, and context lengths, with helpful conversion information about tokenization. 📊

AI Native Developer Tools Landscape - A great resource listing various AI-focused developer tools available at landscape.ainativedev.io. 🛠 The real challenge isn't cataloging these tools but determining which ones actually deliver productivity gains versus those that merely add AI buzzwords to existing solutions. 🔥🔥🔥

Vector-Admin - A universal UI tool suite designed to help users easily manage vector databases like Pinecone, Chroma, Qdrant, and Weaviate without needing complex API calls. 🔍

MCP Server for kubernetes management commands - A project that provides an MCP server for connecting to and managing Kubernetes clusters through your current kubectl context and run commands like listing pods or creating deployments

Software architecture tools - This webpage provides a collection of modelling tools for software architecture documentation. Not much AI yet.

Yet Another AI Research Assistant -An open-source AI research assistant that combines search engines, web scraping, and language models to perform iterative, in-depth research on any topic, with a focus on maintaining a simple implementation. 🔍

Agentic-CursorRules - A Python tool that helps manage multiple AI agents in large codebases by partitioning the file-tree into logical domains (like frontend and backend) and generating domain-specific rules to prevent conflicts.

🎭 AI Humor

https://x.com/bibryam/status/1903802159462076858

📚 Learning Picks

📚 GitHub - mlabonne/llm-course (48K⭐️) - A free course on Large Language Models divided into three parts, offering roadmaps and Colab notebooks, with additional resources including an interactive assistant and a companion handbook co-authored with Paul Iuzstin.

👋 See You Next Week

Thanks for reading Generative Programmer! 🙏 If you like the content please support the newsletter by sharing with your friends via: 🐦 Twitter, 💼 LinkedIn, 🦋BlueSky, ✉️ Email.

Missed the last issue? Catch up here 📖 Got a topic you’d like me to cover? Just drop me an email!

Have a great rest of the week! 😉