Posts

This post continues the idea of Full-Loop Agentic Development (FLAD). The previous article introduced FLAD as a development model where the agent owns the entire loop: planning, implementation, testing, validation, and only then handing control back to the human. This time, we’ll keep it short and practical. Below are two copy‑pasteable prompts, that you can use directly with a coding agent. The first prompt generates a production‑grade AGENTS.md for an existing project. The second prompt kicks off a new project in a FLAD‑native way, using existing ecosystem bootstrapping tools. Both prompts explicitly require the model to research current‑year best practices, tools, and libraries before producing output. ...

Disclaimer: The information in this post reflects the state of coding agents today. They’re evolving fast — what’s hard now might be trivial next month. The Cliff of Death If you’ve spent any time experimenting with coding agents, you’ve probably felt it — that point where everything seems to work until it suddenly doesn’t. The agent writes code, deploys it for you, maybe even fixes a few things… and then progress falls off a cliff. The codebase starts to decay. Small shortcuts compound into big problems. And instead of helping, the agent begins to quietly break things. ...

In an earlier blog post, I asked: Should you version control your prompts? It was a useful question at the time, but today I want to shift the focus. Perhaps it is not your prompts that most need version control. Instead, it may be the context. Why Context Matters When generating code with AI, the context is everything. Context determines what the system can see, what it can remember, and what it can base its reasoning on. At some point, though, context begins to slip out of control: ...

I like straight talk. The hype is real. Investors are searching for the next unicorn. Money flows, demos look great, and some platforms can produce working apps faster than any human team ever will. But if the goal is reliable software that survives contact with real users, real data, auth, policies, and deployment, the road today has a Cliff-Of-Death on it. You don’t fall right away. You sprint, you proceed, you smile, and then — in one final iteration — game over. ...

I’ve been thinking a lot about how we evaluate coding agents, especially after my last post on measuring coding assistants. The usual benchmarks focus on pass/fail results—did the agent solve the problem or not? But the more I tinker with these systems, the more it feels like that lens is too narrow. It ignores how the agent gets there. Imagine two agents. Agent A spits out code that compiles and passes tests on the first try, but the code is messy, brittle, and hard to maintain. Agent B, on the other hand, starts off with a few errors, but quickly cleans them up across iterations, and the final product is well-structured and robust. Traditional benchmarks would rate A higher simply because it “passed.” But in reality, B might be the more reliable partner in the long run. ...

In last years, we’ve seen many tools try to make software creation easier. No-code. Low-code. Drag-and-drop builders. Big promises. But for many creators, results have been limited. Now a new type of tool is coming — AI-assisted Code Generation. It sounds similar, but it’s very different. Let’s break it down. No-Code: Fast but Limited No-code tools let users build apps without writing any code. Usually, this happens through visual builders — drag-and-drop components, predefined templates, and menus. ...

In the past 12 months alone, we have seen massive investments flow into AI-powered coding platforms. Cursor, one of the frontrunners, reached over $100 million ARR with millions of developers onboard, while Lovable signed up 30,000 paying customers in less than a year. These are not flukes or hype-driven spikes; they are strong signals that the market is undergoing a fundamental shift. What Is Driving This Market Shift? For decades, building software has required expensive, specialized human talent. Now, with advances in large language models (LLMs) and multi-agent AI systems, AI can plan, write, test, and deploy robust code from a simple prompt. This evolution collapses traditional development cycles from months to days or even hours. ...

A quick disclaimer: this topic might not seem directly related coding agents, when in fact it does - quite a lot - coding agents are all about prompting. And even if it wasn’t such a critical thing in coding agents it’s something that I’ve learned while working on them and wanted to share. At first glance, instructing a large language model (LLM) with something like “don’t do X” seems straightforward. But my own experiences—and plenty of research—suggest it doesn’t always pan out. Surprisingly, negative instructions often confuse LLMs or simply get ignored, whereas turning these instructions into positive directives makes all the difference. ...

I’ve been thinking a lot about how to make software creation easier—more intuitive, more efficient, more aligned with how people actually work. That’s what drives Gadlet. But if our goal is to empower non-technical users (and businesses) to build software, could the chat box be holding us back? 🤔 Why Chat Might Be the Wrong Default Chat UIs mimic conversations—easy enough for quick questions. But building software is not just chit-chat. Gadlet’s users want intent-driven creation, not dialogue. Chat interfaces: ...

Meet Mikko and Rolf – two Finnish founders with long business careers behind us and one big idea ahead. We are not marketers or growth hackers. We are builders. We’ve worked in many kinds of roles, seen what works, and learned even more from what didn’t. We all know why coding is hard and why you need a required skillset and knowledge to build quality software. Gadlet is our answer to a these questions: Why can’t anyone build software? Why is software development so hard? Why do most great ideas never see the light of day? ...