AI has changed the pace of software development. “Code is cheap” is a statement I think about most days now.

It’s possible to make large bets on technical problems that previously would have fallen straight into the “too hard” bucket. Entire prototypes can appear over a weekend. Complex ideas that once died in backlog grooming now at least get a chance to exist. While code has become cheap, the operational overhead around it hasn’t. The number of signals flowing through my day-to-day work has exploded proportionally.

Merge requests that used to arrive a few times a week now appear multiple times per day. Conversations that once ended with “we should investigate this sometime” now end with a working prototype by the next morning.

“Hey can you review this?”

“I got Claude to generate some tests, can you sanity check them?”

“I made some big DevOps improvements over the weekend, can you make sure login caching still works?”

“This feature had unexpected issues when it hit production data, drop everything, this is SEV 1”

None of these are unreasonable in isolation, but in concert, they lead to constant context switching and inevitably, things get dropped. I’ve always prided myself on my word. If I say I’ll do something, I usually can’t get it out of my mind till it’s done. So, when I was asked for the 3rd time in a week: “Hey, where is this up to?” I had completely forgotten I’d put my name next to it. I knew this problem was only going to get worse; and as most engineers with a problem, i decided to build some software.

The problem statement and original idea were simple enough: “I need a tool to help keep track of all my signals”.

“Hey, can you code review this?” — signal.

“Hmm, this fixes the problem, but we need another ticket for the follow-up work.” — signal.

“Evan, can you follow up that action point?” — another signal.

AI tools and agentic workflow had only just become a topic that everyone was talking about and we had just done an AI workshop to go over the tools now available to us. This had my mind in experimentation mode with the new tools and see what else was available.

I used ChatGPT together with a spec iteration tool called Traycer to create a feedback loop around the product design. I could rapidly explore ideas, refine workflows, challenge assumptions, and expand features faster than I realistically could before.

Very quickly the project evolved into; AI Summaries, integrations with JIRA, teams synchronisation with daily digestion of unread messages and automatic update of existing signals in my new tool. Notion? sure why not. How do i manage the rules? I’ll build an inference rules engine that lets me twist knobs, pull levers to get exactly what i want. AI was extremely good at justifying complexity.

At the same time I was experimenting with google stitch and refining my problem statement to “reduce the mental load” through design while overengineering the core concept and unconsciously building a time bomb that would have killed this project within a week of saying “start implementation”.

Fortunately it was frustration in google stitch not doing what I wanted that gave me a moment of clarity to step back and evaluate what i was making. I looked into the future and saw questions like:

“How do I deal with duplicate signals?”

“How do I test the integrations work?”

“How do I deal with authentication expiry in the various integration points”

I work professionally on enterprise software that integrations B2B and have been in involved in enough support issues to know the value that these integrations give also come with a cost and this project, as currently designed, will only add to the signals in my head rather than help organise the for me.

Around the same time i saw one of those “Make it first then make it good” posts where people chasing perfection were likely going to just have a graveyard of half done side projects. I made the best decision for the projected and locked V1 to be “manual entry only”. Get the simple idea working first, then layer on complexity later if the need actually arises. The app is now optomised to be very quick and easy to create a signal and add a note to update it. Small sentences and an event log when links or notes are added and when the signal is “resolved” that’s it.

“This is my to-do app, there are many like it, but this is mine”

I’m not someone that that thinks vibe coding a to-do app is going to change the world. Its not even an original idea in my own team. But there’s an intense feeling of satisfaction in building something genuinely useful that I enjoy using every day. Its becoming a grounding tool, something I can always go back to when I’m not sure what to pick up next. It feels good to be the one reminding people 2 weeks later because Its still on my list of signals.

The real surprises are how useful the simple features are, having a way to link related notion documents, teams thread, JIRA tickets and GitLab merge request links saves me time everyday. Having an automatic daily stand up sheet generator to give a real representation of the work I’m doing is very satisfying. Having an event history to answer “What’s going on here?”, invaluable.

I have now used this app for just over a month and can see the real value has nothing to do with integrations and things magically happening behind the scenes without my intention, i have abandoned the idea of integrations entirely now. The real value is presenting the information i gather in an easy to digest way. It also gives me opportunities to track the kind of work i do.

My final thoughts - I learned, or I should say, re-learned the value in reducing something to a core problem. AI is an incredible tool for exploration, it reduced the friction involved in brainstorming, spec refinement, prototyping, and even writers block. But it is also incredibly good at helping you confuse designing software with delivering value. I am glad it gives me capacity to explore things I’d have no time for before and I am grateful this particular project helped me refine my AI prototyping process early so future projects also have as much chance for success.