Literally what the thread is about π
Literally what the thread is about π
PixelWatcher is a website monitoring service. Give it a URL, describe the change you care about, and PixelWatcher notifies you when it happens. As a beta user, youβll get a free account with generous limits.
Register here for early access:
4οΈβ£ Recent AI progress is both amazing and scary. I do worry about the broader societal impact, but Iβm choosing to see these tools as an opportunity. If weβre entering a period of creative destruction on an unprecedented scale, Iβd rather be on the creative side
3οΈβ£ The website-monitoring landscape is competitive. Many established players already occupy the space. But that's how PixelWatcher was born β from the frustration of using them. Too complicated, too expensive. I'm building the product I want
2οΈβ£ A lot more people are building software right now. But that's still a tiny fraction of the population, and there's a huge gap between running a prototype locally and launching a real product in production. Very few will get there
1οΈβ£ I don't believe in the SaaSpocalypse. A certain type of SaaS is indeed in trouble β the one with crazy-high margins. But there's room for products that strike the right balance between price and convenience. That's the balance PixelWatcher aims for
On the one hand, the SaaS model is dead. On the other, everyone's building one. Whichever narrative you subscribe to, this might seem like a strange time to build PixelWatcher. So why am I doing it?
A few reasons in this short thread β¬οΈ
Np! Happy to chat
Oh and Iβm using a combination of Claude Code and Cursor (and a sprinkle of ChatGPT for research). I specify and plan, Claude does the leg work, I review and fix in Cursor
To answer your questions though, I verify the output through testing (manual and automated) and I review all code anyway. And yes, Iβm mostly prompting and reviewing, and what really surprised me is how quickly I got used to that. I donβt dislike it, itβs just a different approach
But my guess is using Claude in that context wouldnβt have been as beneficial. Building a personal project from the ground up is a different story though. PixelWatcher is AI-first from the get-go, with the documentation and context this warrants
I think it depends on what youβre building. I havenβt used Claude for client work yet (at the end of my last contract, I had started using Cursor, but it was still very much "fancy autocomplete" then, and agents were in their infancy)
For teams, I agree. Itβs still very messy out there. But as a solo-builder, I can promise you the productivity gains are here (and yet my setup is pretty rudimentary). The models are already good enough β the real bottleneck is the way we organise work IMO (this will take time and break a few eggs)
Fair enough! I meant sustainable in the sense that I see those tools becoming an integral part of the job in the near future, because ignoring their productivity gains wonβt make economic sense. On the other hand, thatβs the beauty of personal projects β you can build them however you want π
Oh wow, that much pressure? Genuine question: do you think that stance is sustainable long term? Or is it also your personal belief that these tools shouldnβt be used?
I understand that the Cursor team is moving fast and shipping a lot, but having to constantly restart the IDE to apply updates is really annoying
π’ Bristol! I'm speaking at PHPSW this Wednesday (11 March).
Iβll be giving an extended version of my talk Escaping the Code Maze, introducing the Closed-by-Default Principle.
Are you based in Bristol or in the area? Come say hi!
Screenshot of a terminal running Claude Code. It just completed a body of work β I'm asking it to commit the code and open a PR, which is met with a "You're out of extra usage" message
Goddammit Claude! What am I gonna do, open my own PR like a peasant? π
Note that it's token-intensive though! Ate through my allowance at lightning speed π
Just tested Claude Code's recently-added `/simplify` skill on PixelWatcher's code (Laravel codebase). Pretty impressed with the result β it found some non-trivial things to improve. Definitely adding it to my workflow
Surely some Canadian is going to pick up on that last sentence at some point
π
Just wrote my first custom PHPStan rule... except I didn't. Claude did. All I asked was "How can we enforce this rule?" Claude was aware of PHPStan in the project, so went ahead and wrote the rule, and tested it by introducing a temporary violation. What a time to be a developer
Itβs on everybodyβs mind, for better or for worse
No pressure! Would just love to meet local developers π
Not sure youβre part of that group too, but Iβll be speaking at PHPSW next month π
Iβm not there yet! But I foresee something along those lines might be necessary down the line, as I add more and more context
Skill to create skills github.com/osteel/.clau...