iiinova — where your digital self keeps evolving

Claude is expensive and slow. Companies are running into this problem where they’re inserting limits and increasing costs when all they needed to figure out is how to use smaller models to get more of the work done. Example: I’m using open source LLMs to write most of my code…

000

Yeah, tests are definitely a great way to help fill the gap. But, there’s definitely a lot more to do to fill it with higher quality output

010

BUT, business logic, authentication, security, SEO, user understanding, and other tooling is now where my attention needs to be.

000

Understanding code itself has provided me huge amounts of leverage: where to worry, where to not worry. For example, frontend code is almost an afterthought since AI can do most of the work (once you lay out the tracks for it).

100

React and Tailwind are one of the best combos. There is so much public code that has been trained on, plus, React and Tailwind both make frontend work super repeatable. 3. A great institution and understanding The more that you can fill the gap, the better you'll be able to leverage AI.

100

Example: buttons. Typically, you can create several variations of them that fill in 100% of the use cases across an entire repository, leading to a strict, yet freeing system to have AI leverage.

100

2. Great design frameworks. What do I mean here? You ultimately get to decide where flexibility exists and where it doesn't. The more strict you make a system, the less of a gap AI has to fill. Sometimes, freedom makes a whole lot more sense than when to constrict.

100

Getting code right in the first stages of a build is where most of the scalability lies. Getting the: - Logging systems right - Spec test structures set up - Error handling correct - Patternistic behaviors you want to use defined in md files These will all help AI build better.

100

I want to talk through each of these areas to help increase your understanding of when and where things fail, and where they succeed. Here's the first one: 1. Foundational programming.

100

There a few ways to get around these problems, and to do so you have to get more specific on what can and won't scale. A great programming foundation scales. A great design framework scales. A great institution and understanding scales.

100

It takes work that's been done, and fills in the blank to the desired outcome you've described. The better you can describe it, the less gaps it has to fill. The better you can do the initial work, the better the quality of the "filler".

100

Ben @benzend.bsky.social · 22/12/2025

I've used AI to build out 80% of the software I wrote in 2025. Building the wrong way can lead to tremendous amounts of problems: 1. Frequent bugs 2. Unscalable designs 3. Broken features Simply put: Great code scales, bad code does not. Why? AI is a gap filler.

211

Philosophy, specifically stoicism, has probably been the most influential thing in terms of changing the course of my life for the better.

000

Build on top of quality. Good software is more important than it ever has been before.

000

And, the more you think about everything you do, in the sense of multiples, you can scale it outward, with almost infinite limits. If you scale on top of slop? That's exactly what you'll get in return, but at scale.

100

Scale is everything. AI is what fills the gaps that were previously near-impossible to obtain. You build not for once, but for many. That one page you built in JS, can now be leveraged to feature implementations, new templates, PRDs, tests, NPM packages, frameworks to be used elsewhere.

100

Good software scales, more now than ever. - It takes me under a minute to spin up several working Figma plugins. - Hundreds of marketing templates within a day. - Several new features in a span of an hour. The catch? It takes the work to get there.

100

Building without having a structure in place will shoot you in the foot, especially if you're using tools that get you closer to developer workflows, like Cursor, Copilot, etc. Shoot to fill the gap! AI tools aren't ready to replace anyone at this point, so it requires the discipline to fill it.

000

5. Review - We need review TWICE in here. One for reviewing the plan, now you get to have the AI review the implementation.

100

3. Revise - Make sure the AI keeps the top 10 security issues, SOLID, DRY, and ETC (easy to change) in mind during the planning and build phases. (will share the file in the comments). 4. Execute - Let the AI move forward once everything looks good to go.

100

1. Plan - Let the AI write out the plan first. Have it build out a plan doc, make sure that it knows it's not writing for a developer. 2. Review - Check out the plan. Make sure that it makes sense. Ask questions and revise if it doesn't.

100

Here are a few rules using a PRRER method (yes, I made that up, but hopefully it'll be easier to remember. Also, it's a rolling RR, like in Spanish) that anyone can take to build software with AI, the right way.

100

What I know is: If done wrong, there are hundreds of traps you can fall into. And this is exactly why experienced engineers are laughing at you while you attempt to build something that ends up not working.

100

Building with AI is f*cking hard. I've been building software on a nonstop, 12+hour everyday grind for the last 5 years (including learning on weekends because I LOVE it).

100

Helping people is important to me. Feel free to follow to learn more about the rules of business coming from a programmer's worldview. Learn to understand your world, instead of being a victim of it.

010

Rules are on a company to company basis, however their are a lot of CORE rules to understand. Life is a bit more complex (and, can even be more fun) to only have one or two rules.

100

3. Research, research, research. You need to understand a company. If you don't care about them, they're even more likely to not care about you.

100

1. Leetcode is the name of the game. GRIND IT OUT. Spend 10 hours a day on it if you have to (or can). 2. Learn their stack. Get so good at it you can spin several applications up in a day.

100

How do you know the right areas to grind? Understand the rules of the game. Look at them logically, without hate, and allow yourself to follow them (or break them entirely). Programming is SOOO similar. Looking to work at a big company? Here are the rules:

110

Look at this like a Dark Souls game. Need to beat the next boss? You'll have to grind an area for larger amounts of souls for leveling up. Or, another area to get smithing stones to improve your equipment. It all comes down to "time-grinded" in a dedicated area.

100

I want to share this simple rule. TIME IS THE ULTIMATE LEVER. Time is what allows you to become the top 10%, top 1%, top 0.1% programmer. Leveraging it will get you to where you need to go.

100