I have absolutely no business being here.

My background is small business and government contracting. I ve never been a software engineer. I don t have a computer science degree. Honestly, a year ago I probably would ve panicked if someone opened a terminal in front of me.

Now I m sitting here at stupid o clock in the morning, building an actual SaaS application with AI. A real, multi-stage thing with logic, workflows, databases, testing, bugs, more bugs, and an alarming number of conversations with Claude that basically amount to: Okay. That broke something else. What did you do?

We've been getting the same request over and over from our users: "My AI gives great answers, but it can't actually do anything."

It got us thinking most AI integrations today are still essentially fancy search boxes. The AI talks, the human acts. But the real unlock is when the AI can close the loop itself query the database, send the email, update the record without a human in the middle.

The hard part isn't the action itself. It's the non-determinism. How do you build a system where the AI decides when to act, which action to take, and what parameters to pass based purely on context without it going off the rails?

A few things we've learned building this:

Last weekend I tried Lovable to build a simple habit tracker, and the starting results were mind-blowing. But by Thursday, every small tweak started breaking existing features I wasn't even working on. Prompts became exhausting, and my patience was tested.

I started attaching a small constraint document: a brief summary of the app's goals, explicit restrictions on what it shouldn't alter, and a record of all error-free screens. I paste this right at the start of every big prompt.

It s far from perfect, but it reduced unexpected code breakages significantly.

Do you paste guidelines with your prompts, or do you rely on starting from scratch each time?

A few months ago I made a small bet with myself: build and ship a tiny tool in one weekend, no overthinking, just vibes. I did it. It worked. People used it. And then something strange happened.

The weekend project taught me more about my own habits than any "proper" project ever had.

I fix WordPress and WooCommerce problems for a living, and AI has changed some of my routine. It's very good at reading an error log and explaining a stack trace in plain words. I still don't trust it to edit a live checkout or a plugin I don't know well.

The places it has sent me the wrong way are usually plugin-heavy sites, where it suggests a clean fix that ignores another plugin hooking into the same function.

I'm what you might call a gringo vibe coder.

No dev degree. No traditional background. Just me, Cursor, Claude, and AI... figuring things out together.

Rick Rubin said it best: "In the past, for music, you had to go to the conservatory and study for years... then when punk rock came along... if you had something to say, you could say it." He calls vibe coding the punk rock of coding.

about to Launch a product for creators after building for a few weeks to months.
Started as a way to get into vibe coding and just got real useful real quick lol.
But the goal is still that - mix my skills and knowledge as an engineer with the good parts of AI coding.
So i learned a few of what to need and have

• Claude.md file

• best coding practices section

• pretty heavy compartmentalization & file structure separation

• certain ways to prompt

• ask for implementation plan before telling it to code

But i feel like all i learned is about directing prompts - im wondering...

what are you guys finding are the best .md or deeper ai vibe coding tricks that are helping you save time? be more efficient? debug ai less or help it get it right the first time? whats your vibe code tips and tricks?

Been meaning to do this properly instead of just going off gut feeling every time my usage bar turns orange. Pulled together what I could find from benchmarks and my own week of switching between tools, and a few things genuinely surprised me.

The efficiency numbers aren't close, and not in the direction I expected. One widely cited test had Claude Code completing an identical multi-file feature task using around 33,000 tokens with zero errors, while Cursor Agent burned roughly 188,000 tokens on the same task and hit several errors along the way, a 5.5x gap for the same deliverable. Separately, other community head-to-heads put Claude Code at roughly 3-4x the token use of Codex on comparable work, so depending which pair you're comparing, the "efficient" tool flips.

Zero-shot capability and token frugality don't always come from the same tool. On raw benchmark performance, Claude Opus 4.6 sits around 80.8% and GPT-5.5 powering Codex around 88.7% on the metrics one source tracked, but Claude Code holds an 88.6% SWE-bench Verified score by another measure. The benchmarks genuinely don't agree with each other depending on methodology, which is its own lesson.

Price-per-output tells a different story than raw capability. Cursor's Composer 2.5 lands in the top three on capability index while running roughly 10-60x cheaper than the leaders, so if you're optimizing purely for $/task rather than fewest tokens per task, the ranking flips entirely.

I counted the tracked files in my main repo this morning, expecting the code to dominate.

280 markdown files. 123 React components. The markdown runs 427,099 words. The components run 23,394 lines.

I have been trying vibe coding more lately

Its crazy how fast you can go from an idea to something that actually works

But getting the first version out is easy

Making the code clean enough to keep building on is a different story