Daniel Philip Johnson

Daniel Philip Johnson

@daniel-philips-enterprise.com

Fintech developer 💼 | Frontend architecture with React & Vue & Next.js ⚛️ | Golang & PHP 💻 | Focused on chat & security OWSAP 📝 | Bonsai enthusiast 🌱

400 followers451 following1.5K posts

Currently rewriting lots of internal web apps from create react app to vite. Is it worth writing a blog about it all. So far biggest impact before to build on a small project was 3.9s it’s now 0.25s. Bundle size saving about 12% on average.

AI hasn’t eliminated the attack surface it’s multiplied it. Now companies are deploying the same agents that can infiltrate codebases to defend them. We’re creating new vulnerabilities, then using AI to protect against AI. Progress, or just an arms race with better marketing?

The patterns I’m starting to see. The code is over defence but at the same time add millions of fallbacks instead of making bad state possible. Creates bad abstractions. Avoid strong invariants. The less humans are in the loop the more it’s going to compound.

Pay lots of money to get AI to push crappy code? YES!! Pay for visual regression? NO!! Too expensive!! Invest time into pipeline to make sure AI code is governed. No!! Makes merging to production slower. Make it faster to push code to production by removing stuff from the pipeline? YES!!

Ironic watching Microsoft build goodwill through VS Code and GitHub, then start rug pulling trust in the AI era. Google too: years building strong developer ecosystems, now “progress at all costs.” Everyone racing to be first to AI(LLM). Stability and trust become optional.

Jake Archibald@jakearchibald.com

So, Chrome's "web standard" Prompt API: Mozilla: Opposed WebKit: Opposed Microsoft: Several concerns W3C TAG: Several concerns Developers: Mostly negative Chrome: Ships anyway. A sad time for web standards. But, I guess someone at Google will get promoted, so 'every cloud…'

One of the most common office dynamics is when someone asks for solutions to a problem they’ve already decided is impossible. “We need to reduce bugs.” “Maybe we improve test coverage.” “Yes, but that slows delivery.” “Maybe we tighten ownership.” “Yes, but teams are already overloaded.”

“Maybe we simplify the system.” “Yes, but the business needs flexibility.” At some point you realise the conversation was never about solving the problem. It was about protecting themselves from responsibility. Because if every solution is flawed, then nobody can blame them for failing to fix it.

Greenfield systems built primarily with AI assistance in 2024–2028 will, in 2035–2040, present the same archaeology problem that COBOL systems present today: large volumes of functional code whose internal logic is opaque, whose authors are no longer available, and whose documentation is inadequate.