Peter Prins @PeterPrins10

Peter Prins@PeterPrins10Engineering · #5

I've been building AI apps completely wrong. 🤯 Everything my app needs from users already lives in their ChatGPT or Claude: their research, their drafts, their files, their context. I spent months building onboarding flows and import UIs to collect data users had already created one tab over. 🫣 Every UI step is a drop-off point. Every form is a reason to churn. 📉 The fix: a well-written skill + an MCP server. The user's AI already has the context — you just give it a structured way to hand it over. 🚀 Shopify gets this. Every store now has an MCP endpoint by default. A merchant who researched products, wrote descriptions, and generated images with AI shouldn't re-enter any of it — the agent builds the product page. Don't build another import wizard. Build the skill.

Use MCP/skills instead of import wizards in AI apps

Peter Prins@PeterPrins10AI · #4

I asked both Fable and Sol to work on the same code base. Then I had both Fable and Sol review each other's code blindly. The results may surprise you. t.co/4xn6XVGkpY

Methodology for comparing LLM output quality blindly

Peter Prins@PeterPrins10Product · #8

Using Claude, we built this animation directly from our codebase and the result is pretty incredible. Simply helping users create is not enough. For AI companies to succeed we need to actively help users sell their products. This is the direction we’re taking with Vellum. Animations will be a big part of that. They will help our creators turn their Instagram and TikTok followers into paying customers. I'm very excited about this, what do you think?

AI product success requires helping users monetize

Peter Prins@PeterPrins10AI · #6

LLM model switching comes at cost. Every model LLM has it's own quirky outcomes. Learning these quirks and adjusting your instructions takes time. Therefore, there is a real cost to switching models. The code below is written by GPT-5 on Auto with the new Codex plugin. t.co/HCcc442n3q

Explains hidden cost of switching models

Peter Prins@PeterPrins10AI · #7

Fable is the last model you need for programming. We don't need the model to be smarter. We need it to be less token hungry. There should be enough headroom for Anthropic in post-training and the system prompt to make the model perfect.

Token efficiency matters more than capability for coding

Peter Prins@PeterPrins10Product · #13

How can Replit be such a valuable company? It's just a wrapper around Google Cloud and some other services like Neon. Even much of the UI is borrowed. For example, it's using Drizzle Studio in it's UI for to view the database.

Company value comes from UX integration, not just tech

Peter Prins@PeterPrins10AI · #8

Using Grok 4 Fast today - So far it's a better experience than GPT-5. Unfortunately, it's not implemented in Cursor just yet. So you will need to set it up manually and use your own API keys.

Teaches setup method for using new LLM

Peter Prins@PeterPrins10Workflows · #11

Cancelling my SaaS subscriptions. Today I cancelled SuperHuman as I found Claude Cowork finally connects with Office 365. Now I will try to automate my email workflow and calendar scheduling with skills.

Claude skills replace SaaS tools for email automation

Peter Prins@PeterPrins10AI · #22

Is Apple the secret winner of the AI revolution? Closing a billion doller deal with Google to create an Apple specific Gemini model, may be the most cost effective way to bring state-of-the-art AI to Apple devices without huge CAPEX investments.

Licensing AI more cost-effective than building in-house

Peter Prins@PeterPrins10Engineering · #13

I spend the last week using the Codex app to write software. I love it. It improves my productivity over working in terminal. Things I love: 1) It's easy to launch tasks in separate worktrees to work in parallel 2) Easy access to code changes 3) Integrated terminal, that opens inside the worktree where your task is working. 4) Clear overview of all your runnings tasks (albeit this could be improved) Compare this to working in the terminal with CLI tools: switching back and forth with your code editor to see code diffs, needing manage multiple windows / tabs to access terminals. Needing to the switch to the correct worktree. The Codex app lightens this work for me, and allows me to focus on the next task. And yes.. the Claude app is similar, but GPT 5.4 > Opus 4.7 + Codex app > Claude app (as it still has a lot of bugs). May give Conductor or T3 Code a try next week.

IDE workflow advantages for parallel task management

Peter Prins@PeterPrins10AI · #21

Programming with AI is full of surprises. Sometimes it astonishes me. Complex tasks, perfectly one shotted. Great code. Then I ask it to do something easy, even for a junior dev, and the code AI generates is just horrible.

AI code quality is unpredictably variable by task type