Peppe Silletti

I’d love to see more people sharing their stories about solving product challenges. We often get caught up in evangelising tools, frameworks, or methodologies in isolation, but what’s actually valuable is understanding why and when they work well.

How to unf*ck the "what are we even building”? The work on Vera Field continues, and after many discovery calls, we got a good overview of all the challenges field technicians face when creating service documentation and how we can support them.

Discovery calls gave us so much valuable information, and my brain now screams for order, structure, and a hint of what's coming next. I need to see the picture on the puzzle box so I can start assembling the pieces one by one and make sense of everything.

peppesilletti.substack.com/p/how-to-unf...

Building a Bootstrapped B2B Product with Gen AI (0 to 1) Follow along as I go from ICP → recurring, unmet need → solution-idea → assumptions → testing → sale for a real B2B product. My own.

My co-founder talks about it from her PM point of view here: open.substack.com/pub/elsevand...

"Want to build something?" Building a bootstrapped micro-saas and experimenting with V0

Documenting my journey co-building a bootstrapped B2B micro-saas : open.substack.com/pub/peppesil...

6/6
Find it here: www.peppesilletti.io

5/6
Tech Stack:
- Astro
- Tailwind
- React
- @portaljs/remark-wiki-link
- MDX

4/6
- Have a way for me to modify the content from Obsidian, and so I had to find a way to parse Obsidian's wiki-links (used @portaljs/remark-wiki-link for that)

3/6
- Create a section to show all the communities I've founded / belong to. I called it "Campfires"

- Show backlinks in every content item's page, which represent the connections between the whole content items collection

- Let visitors subscribe to the website RSS feed

2/6
- Offer a way to group related items, and I thought about introducing "trails". For example, all the episode I've recorded for Invisible Algorithms belong to the "empowering software engineers" trail

1/6

Building this make me feel like Iron Man with Jarvis (I wish, I just used v0 a lot). My idea was to:

- Create a way to show all my notes/essays/videos/podcasts in a single place in a random order, and offer some tools to navigate them. I called it "The Lookout"

Finally had some time to build my own version of my Digital Garden.

The concept of #digitalgarden really suits the way I think and approach content: not a series of time-ordered blog posts, but a network of non-necessarely-completed items that grow over time and are interconnected to each other.

Every time I open the code editor—poof!—the name of the person who wrote a line of code pops up, reminding me that every decision I take today will be written in stone. I'll be judged and damned for eternity by every developer that'll come after me.

2) Generate the scaffolding for new features

When your codebase is modular, it becomes easier for Cursor to understand the patterns its using, and to generate new code that fits that structure.

Careful: if your codebase is messy, Cursor will try to fit the same style, making it even messier.

1) Exploring a new, unfamiliar codebase

When you're working with a codebase you're not familiar with, and you want to edit a feature, you can ask Cursor to find all the files that enable that feature, and describe how they work together.

Then you can ask for some tips on how to edit it.

My two favorite ways to use Cursor to speed up my development:

1) Exploring a new, unfamiliar codebase
2) Generate the scaffolding for new features

👇

Engineers who think beyond code: welcome home. We launched a community for product engineers!

Share knowledge, discuss practices, and grow together.

No gurus – just practitioners helping each other build better products.

Ask to join here: z0xsrwr8ltf.typeform.com/to/Y5yUYrif

Just want to publicly thank @maggieappleton.com for making it possible to discover Digital Gardens through her article maggieappleton.com/garden-history

It gave me hope for a better way of sharing content online.

I got into a rabbit hole and came out with my digital garden V1: peppesilletti.io

🙏

Let's say I build an entire software system all by myself. It goes into production. It works fine. Quality is decent. The company makes money.

Question: Is my work considered less "software engineering" compared to the same system being built by a team of 10 people?

I know how to use a brush, that doesn't make me an artist.

I think the same will be always true for all the AI tools assisting with coding tasks.

So you're going to be a winner.

Can I bet somewhere?

One thing is true: it's more fun looking for solutions than just implementing them.

That explains all the noise of the last days...

I haven't read the report, and I should. But question: did they take in consideration what tools the devs used?

There's a huge difference between copying pasting from the chat, Copilot or Cursor for example.

Maybe we just need less developers in a team, easy as that? Less people, less coordination.

Jason, I'm in a real rush to finish a project and I'm really going fast thanks to Cursor, without sacrificing quality.

Why would I ignore this for the sake of protecting the team from bottlenecks?

Pivoting how? Learning to integrating it in software systems or actually building the behind the scene?

I am my team! :)

I get what you mean though. Let's say there were multiple *me* working as I do in the team (in isolation or pairing/mobbing), why would the help I'm getting from Cursor be less valuable?

In my specific use case it really helped me speed up my work (refactoring an old code base).

I use Cursor, and once it starts to understand how I work and the project's patterns, it starts to give great suggestions.

Not sure how the report was made, but as always, context is king.

Are forums still a thing? Really miss them.

The welcome thread...

People actually getting to know you...

A real sense of community!

To give some context, there's already a backend V1, so I'm just rewriting the all thing with a different tech stack.

I use BDD, so I start with an acceptance test and then let Cursor generate all the code to make the test pass.

Or even ask it to generate the test first.

I've been using Cursor since July while building the new Yoop Knows backend. It's been a life saver.

Once the project started to have a defined shape with recurring patterns, it became quite good at generating new stuff or with autocompletion.

I'm going so fast and all I'm doing is reviewing code.

The biggest challenge right now is not to prompt an AI model to produce code. Almost anyone can do that.

The challenge is to validate that this code is actually doing what you want it to do. The challenge is to understand how to improve it, fix it, and update it.