I’m 44. Solo founder / builder from the UK. Looking to connect with more builders & indie hackers. Currently shipping a lightweight cron monitor for solo devs. Always up for a chat with people doing the same.

PulseWatch has a watchdog that checks whether its own scheduled monitoring is still running. At first that felt slightly ridiculous. Monitoring the monitor. Then I realised the worst possible failure for a monitoring product is confidently sitting there doing absolutely nothing. So yes. The watchdog stays.

Apparently 42% of developers now say AI writes at least half their code. I’d comfortably be in that group. But “percentage of code written” feels like an increasingly useless measure of how much of the software you actually built. The interesting bit moved somewhere else.

One of the great injustices of parenthood is finally getting the children interested in something you love, only for them to immediately become better at it than you.

I’ve noticed a useful warning sign with Codex: If I can’t explain the change I want in a reasonably tight prompt, I’m probably not ready to ask it to code yet. Ten extra minutes figuring out the problem usually beats twenty minutes correcting a beautifully implemented misunderstanding.

The Wikipedia experience of looking up one perfectly ordinary thing and resurfacing 45 minutes later reading about a 14th-century succession crisis remains undefeated.

Hey team! A bit of a slow start to this weekend. Kid’s football takes precedence. But now I’m locked in. 3 clearly defined tasks in front of me. Flat white next to me. Let’s freakin’ go!!!

That viral “human meat proxy” post about AI coding struck a nerve. I get it. But my experience building solo has almost been the opposite. Codex writes far more of my code now, yet I’m making more decisions about the product than I ever could when every feature meant hours of typing. Maybe autonomy matters more than authorship.

‘Semi-Charmed Life’ is real wild, eh?

Friday night BBQ. What are we all cooking?

GPT-6 Sol arrived this week at half the API price of 5.6 Sol. I wonder whether model cost is actually a terrible reason not to try an idea. Time, distribution and whether anyone actually wants the thing remain stubbornly expensive.

Wikipedia might still be the best thing the internet ever built. Twenty-five years in and I can lose an hour because I wanted to check one date.

One habit I’ve picked up with Codex: I ask it to leave the repo boring. Use the existing patterns. Don’t add dependencies unless necessary. Don’t redesign something adjacent because you noticed it.

“If you can’t be good, be lucky.” I updated my Cloudflare analytics on PulseWatch this morning. The deploy failed. Not because of my change, but because SQLAlchemy released an update yesterday that switched its default Postgres driver. My analytics problem caught a monitoring problem 😅

I genuinely think “sleep on it” is one of the highest quality decision-making tools available to humans. Hard to monetise.

Coding agents are now good enough that I’m much more interested in what they’re allowed to touch. Shell. GitHub. Secrets. Production. Other tools. Giving an agent more capability feels fantastic right up until you remember that capability works in both directions.

Let’s go team! If you don’t follow…follow @NikChr1414

Every dog owner knows the precise difference between “the house is quiet” and “the house is suspiciously quiet”...right now I’m suspicious.

I wonder how many side projects now exist purely because AI reduced the cost of scratching an itch. Not businesses. Not startups. Just: “That’s annoying. I reckon I could fix that tonight.” I’m pretty sure I’ve accumulated several.

The older I get, the more I appreciate people who can explain something complicated without making you feel stupid for not already knowing it. You know who you are. And I applaud you.

One thing I didn’t anticipate with PulseWatch: building the boring infrastructure is oddly satisfying. Idempotency. Restart recovery. Canary checks. Making sure the thing watching failures doesn’t quietly fail itself. None of it demos particularly well. It’s probably the bit I enjoy most. Certainly beats marketing.