vincent_s

You got that wrong. This is a copy of Figma, not "the next Figma". So it proves the point I tried to make.

That's why I suggested blocking requests at the web server level so they don't even reach PHP.

Two things you can do now:

1. Turn off session upload progress

2. Block multipart requests at the web server

More info: https://www.vincentschmalbach.com/cve-2026-109618-and-cve-20...

Two things you can do now:

1. Turn off session upload progress

2. Block multipart requests at the web server

More info: https://www.vincentschmalbach.com/cve-2026-109618-and-cve-20...

I created explainroo for this. Just point your coding agent to the repository and give it the website of your product and it will do all the rest for you.

https://github.com/vincentsch/explainroo

Absolutely. But "run it yourself" is completely underestimating how hard it is to run inference at scale for large models and at the same time bringing costs down. The experience (speed, costs, cache) I have with various providers and models on OpenRouter is way worse than using Codex and Claude and if you want the outcome to be comparable in quality it also costs a lot more than using OpenAI's and Anthropic's subscriptions.

Btw. I absolutely agree with your analysis that models seem to be more or less plateauing for real world use cases and that OpenAI is pushing them into the "try anything and everything until "the task is complete"" direction. They should have stopped at GPT-5.5 and rather tried to make this one cheaper and faster. The rest is then the job of surrounding tools and/or the "harness". For my SaaS [0] I am still using GPT-5.5 with great success and have no plans of upgrading to GPT-6.1/Sol. GPT-5.5 is good enough and gets the job done reliably. GPT-6.1/Sol is fine as a reviewer of you take it's output with a grain of salt and not follow it verbatim.

[0] https://vroni.com/

I think they try pushing the models in a direction where they just go on until somehing is finished end-to-end. So basically building "the loop" into the model itself. E.g. with GPT-5.5 the model would often do only part of a job and get back to me with a recommendation for next steps and I would either say "Yes, go on" or "No, instead do X". And the newer models never even get back to me, they just keep on churning, deciding about the direction on their own. Sometimes they go into the right direction, sometimes they drift into something unrelated or just go down some rabbit holes for hours. I liked GPT-5.5 better, and I built my own tools around it to still get them to finish tasks end-to-end which works fine.

Location: Germany (CET/CEST)

Remote: Yes

Willing to relocate: No

Technologies: Laravel, PHP, Vue.js, Inertia.js, Livewire, React, TypeScript, Python/FastAPI, Node.js, MySQL, PostgreSQL, Redis, Tailwind, Docker, OpenAI/Anthropic/Google AI APIs

Resume/CV: https://www.vincentschmalbach.com/

Email: see profile

Freelance full-stack developer and AI engineer with 15+ years building web applications. My main lane is Laravel/Vue business software: SaaS products, internal tools, API-heavy apps, legacy PHP modernization, and projects that need an experienced person to stabilize them and move them forward.

I also build AI features into production software: OpenAI/Anthropic/Google API integrations, agent-style workflows, scraping/data pipelines, queues, and content/publishing systems.

Founder/operator background from SaaS, ecommerce, SEO/content, and marketplace projects, so I can help with product scope and business tradeoffs as well as implementation. Best fit: remote freelance or contract work, especially 20+ hours/week or focused rescue/build projects.

Thank you for the kind words, glad you like it.