Old and new processes.

Sometimes the old ways are the right ways.

And sometimes the new ways are crazy.

Topics for this week:

• Webflor Re-Evaluation

• Simple Finances

• Grokipedia

• Micro vs Monolith

• Claude Code Pro

Webflow Re-evaluation

It’s been a surprisingly light week in development.

I’ve had Cursor open a couple times but I spent more time in Chrome on Webflow’s instead.

Making some updates to an existing website made me realize how fast things have evolved in such a short time.

I found myself thinking it would be so much easier to just prompt this change and majority of the work could have been done in a single line.

Initially I started with Webflow because it’s a great service.

They provide a ton of convenient features with domain management, super fast deployments, a CMS, form submission handling, and a good balance of access to advanced CSS and animation settings that can result in impressive looking websites.

But lately it’s feeling like nearly all of that can be handled through prompting.

The infrastructure is still the biggest convenience provided by Webflow, so that’s why I haven’t immediately jumped ship.

But for a simple landing page, I’m not sure it’s worth it anymore.

The CMS is helping, but creating a new Content Management System isn’t too far beyond prompting a database service into existence.

Sticking with it for now, but a Webflow migration may be on the list for 2026.

Simple Finances

Talking about tracking finances this week, I was reminded of the ever-present business problem of “build versus buy".

Is it better to save the development time and cost to buy something that’s close enough to the businesses needs but still requires further customization?

Or is it better to take the time and money to build something that does exactly what it needs to do and hope that it actually works?

In the case of tracking finances, the “buy” option is actually free.

There’s plenty of free apps for tracking finances, and they all have their own advantages.

But there’s something to be said for building things manually on a spreadsheet.

Apps are way more convenient to jump in and get started with pre-determined features, but will for sure not cover every use-case.

Like in Rocket Money, I couldn’t find a way to update the built-in Calendar feature so I could add a yearly fee it wasn’t expecting.

If you don’t have a specific use-case, the apps are good enough.

If you want to do anything outside what they provide, then there’s an issue.

A spreadsheet on the other hand, takes way more time and effort but it makes up for it with complete customization.

There’s a reason spreadsheets have remained such a staple for all these decades despite all the advances in software.

Highly customizable but highly unintuitive.

Plus I think there’s a hidden benefit of spending the extra time to process in your own head what you’re adding to the spreadsheet rather than passively checking a list on an app.

Sticking with the theme of simplifying this year, a big “finances” spreadsheet may be the play instead of using financial apps.

Simple and exact tracking, worth the extra effort.

Grokipedia

Spotted this one on X the other day.

Looks like Grok has generated more than 5 million articles to create it’s own AI-maintained Wikipedia alternative.

I was going to refer to Webflow earlier as a “WSYIWYG” website editor, so here’s the link to that term in Grokipedia: https://grokipedia.com/page/WYSIWYG 

An AI-based Wikipedia alternative is an interesting idea.

AI has no problem generating articles these days, so it can easily create new pages and keep things up to date constantly.

And AI could also help with one of the big problems with Wikipedia in monitoring for false updates.

Since Wikipedia is supposed to be maintained by everyone, anyone can make edits that aren’t true, so there also has to be constant monitoring of any updates to make sure they are accurate.

Grok, or any other AI, can at least monitor the constant updates and do it’s best to verify, or flag the update if it can’t verify the changes.

Kind of surprised I haven’t seen this sooner, but we’ll see if it ends up competing as a go-to reference like Wikipedia.

Micro vs Monolith

LinkedIn suggested a good post the other day and I should have saved the link because I can’t find it now.

But an engineer made a point about the back and forth between software approaches: microservices versus monoliths.

Do you put the code for the application in one big code repository or do you break it up into microservices where each functionality is it’s own service and repository?

These days it’s usually a mixture of both, parts of the app will be a huge repository, then there will be supporting services with their own repositories.

Or there’s one big repository of code, but it’s broken up into separate services.

But in this post, they pointed out that the microservices approach primarily benefits the humans maintaining the code.

Of course, there’s a stability and scalability advantage to microservices.

It keeps services separated so that one service can stop working without making the entire application break.

But microservices also separates functionality into individual pieces so it’s easier for humans to understand what that particular service does without having to read or know the entire application.

It keeps things relatively self contained.

AI on the other hand, loves context and would perform much better if it can read all the code for the entire application rather than having to operate with only the context of the microservice.

Massive repos could become the norm as humans do less and less coding.

It makes sense to me even though I personally like the microservices approach.

Throw all the code in a big repo and let AI sort it out.

This is the future.

Claude Code Pro

Speaking of the future, the bar has been set.

Earlier this week, the create of Claude Code walked through his agentic coding process:

Unsurprisingly he is a power user.

He typically runs 5 Claudes on his computer, while also running 5-10 instances of Claude in the cloud on http://claude.ai/code.

At any one time he’s running up to 15 instances of Claude Opus 4.5 to work on code or do code reviews.

He said his process is surprisingly light, he has a few commands he uses, but generally these agents are all guided by a constantly updated CLAUDE.md markdown file that includes instructions to help Claude do things the way he prefers to prevent mistakes.

I’ve seen others post online about how they use up to 5 agents at time to work on a task, but 15 is a new level for me.

This is something I’d love to try out but the cost is a little scary.

There’s a high-tier subscription for Claude Code Max, but I’m pretty sure you’ll be hitting your $200 per month usage caps with 15 agents at time after a few days.

He might also be benefitting from working at Anthropic so his usage limits are probably more generous than the paid tiers.

But with just one person running 15 instances of an AI at a time, it’s easy to see why there isn’t enough compute to go around right now.

Super cool to see how the creator of one of the most popular AI tools is personally using the tool he created.

I’ll be anxious to try some of this out myself at a cheaper scale.

But with 15 agents at a time, you can almost feel the AGI…

And that’s it for this week. Webflow, finances, and Agents galore.

If you want to start a newsletter like this on beehiiv and support me in the process, here’s my referral link: https://www.beehiiv.com/?via=jay-peters.

I also have a Free Skool community if you want to join for n8n workflows and more AI talk: https://www.skool.com/learn-automation-ai/about?ref=1c7ba6137dfc45878406f6f6fcf2c316

Let me know what else I missed! I’d love to hear your feedback at @jaypetersdotdev or email [email protected].

Thanks for reading!