Morning and happy weekend everyone

Just rolled out of bed to start chipping away at a core part of the Dateful experience, but first, I will write and humbly share that yesterday was a miss.

A painful and frustrating day that tested my patience with myself and kindness to patient coding companion. But there’s a big lesson here…

I carved out 2 hours thinking I'd get way ahead. My goal was modest and simple: get the app to send a text to my phone, connected the model, one inbound message to me, one outbound from me, holding my couple’s context. Done.

I had all the infra in place from yesterday and a robust plan.
Plus I've done much more in two hours before.
This seemed like it would be easy.

It was not. I spent almost all my time dealing with failures. Calls not working, edge functions returning errors and oscillating between different types of errors, APIs not playing nice. Just copy and pasting problems, and wrestling with Cursor, dev tools, Supabase, and Twilio for nearly two hours straight. Mega frustrating.

I almost didn't write about today because it's so lame to show no progress. But this is the reality. Sometimes there is no difference from one day to the next despite you pushing for it and doing the work.

This is what actually happens. There are peaks and troughs and even though AI is doing the heavy lifting, there's a layer of real patience and debugging stamina you need that isn’t obvious when just watching videos of people vibe coding.

I’ve built things with v0 before—like this set of wealth calculators— that were poorly architected at first (before I knew not to trust a model to code me something scalable and extensible without explicit plans to do so), that led me to massive pits of tech debt where it was impossible to get basic changes done before a focused rewrite. It made me a monolith code base and could not maintain context. Just syntax errors and breakages and failures until I broke it up.

I’ve built apps for IOS and Android—like this personal favorite of mine—that were built and architected fine (lesson learned from above), but have taken hours and hours to just get libraries compatible and something to just render and show.

It happens and even with high levels of planning, and it does suck, and it does deflate. The early part of a new AI coded project often from my experience has a lot of opportunity for bugs and error prone states that will shake out anyone who isn’t prepared for some patience to see it through.

Anyway, I learnt another vector of failure in these projects.
And I learnt something else important.

The big blocker I had: Twilio now requires you to register as a business and go through A2P 10DLC compliance before you can do outbound SMS. I need this so I can message people date plans.

This is relatively new and I did not learn this until about 90 minutes into my agony, and if you're planning any SMS-based product, this is a massive heads up.

You need to register your business entity, then create "campaigns" for each type of message you want to send. Easy, but the approval pipeline is 2-3 weeks. Not days. Weeks. wtf?

If you're going to do this, register campaigns for every message type you think you might need upfront. Don't do one at a time. Get them all in the pipeline together so you're not waiting in serial.

This could put my 30 day timeline at risk. My channel is text and I can't send outbound (agent to user) texts for weeks. So likely will have to adjust here: build the web chat experience first. Not a big deal system wise, it does let me build a solid core app experience first and get the recommendation engine right. But it means the text magic that I think makes Dateful feel like an actual personal assistant is on hold.

I might have to launch without text entirely.

As soon as I realized this compliance issue, I moved on. But I should have earlier than that.

And this is my own lesson for the day. I spent far too long trying to brute force the phone verification flow when I should have just changed my goal and moved on. I was thinking in waterfall—sign up, auth with phone for the texting channel, then connect the brain. Step by step, in order.

But nobody uses this thing yet. I could have cut the phone complexity yesterday and invested that time in the main experience instead. Email auth worked out the box with Supabase which was good enough, but I was being rigid about a sequence that didn't matter yet. I wanted the phone to work really badly because that is a personal goal of mine for this build.

Step 2 was blocked, should have skipped to step 4. Nobody's using the product yet and only fast progress matters on a crunch like this.

Lesson: When you hit a wall, assess if the wall is truly a blocker, and if not, walk around the wall. Screw the wall. Change the goal. Don't keep smashing into it hoping the wall moves. Especially when you're building solo and are the PM, the engineer, and the QA team. Time is the only resource you can't buy more of. Come back later.

I realized this too late, and now with only 30 minutes left to try find a win.

Anyway, in the last stretch of my time box, after I finally stopped fighting Twilio and it’s related functions, I got this part built and verified in my local.

That last one is a good milestone. The model replies. It has context. It works. I just couldn't enjoy the progress of it like I wanted to because I was stuck in a debugging pit for the first 100 minutes.

I was talking to a friend (howdy Joe) who's been inspired by these few emails to start building his idea. We were on the phone while he was playing with Claude Code, saw how much it could do, and told me about what he wants to build. It was a big idea. A great idea. But a huge bite for a first project.

My advice was to get something that does nothing interesting out the door first. Just complete a workflow. Something basic. Go through the motion of a commit and push with a basic basic feature. Because even simple things can trap you in a debugging pit that if you’ve never built before, can wipe your energy.

It's easy to lose motivation early. The goal of the first thing you build isn't to be the thing you want to build. It's to feel excited and inspired enough to build the next thing. Don't start with the big thing. Start with the small thing that makes you want to keep going.

Also, more validation around the premium feature discussion last time. R.e: “watch and book hard to get into places for you” . Definitely something here.
I’ll come collecting soon Ali 😉

See you after some more progress. Hopefully.
— Jaryd

p.s I launched a free forum on my site for us to chat about things we’re building