Here are the tools I use daily at work. All of those are must haves for an AI native startup these days. @cursor_ai @NotionHQ @posthog @perplexity_ai @figma @semrush @Setapp @1Password

Interesting.

Switching from claude code to grok code fast in Cursor. The speed makes a huge difference, even if it is a bit dumber.

Is Opus 4.1 slightly too coding focused now? I asked it to brainstorm and help me make a decision and it created a ridiculous, colorful html artifact straight away, instead of answering in the chat. Maybe the focus on claude code is making the Claude chat app worse?

Having Claude Code do my Devops and SRE work is now called "Flibbertigibbeting…" 😆 What horrors could it have read in the logs?

What nobody tells you about AI coding: the gateway drug progression from Cursor to Claude, why you'll hit rate limits at the worst possible moment, and how 'fake intelligence' can still transform your productivity. Spoiler: You'll end up on the $200 plan https://t.co/YoFZdq8SLZ

This basically replaces all SaaS marketplaces like G2 or Capterra as you can feed all the data into your Claude or Cursor. Your LLM then becomes your advisor. 🔗

I'm giving away a fully enriched Database of all SaaS products, their pricing, G2 reviews, Trustpilot, feature info, alternatives and competitors, growth metrics and funding. Rason: we're pivoting to a slightly different business.

Link to stack 👉 https://t.co/yhh7vJjrtV You can now easily make money by sharing your own stack!

this is how i code now

I see you vibe-coding javascript games from and raise you vibe-coding SaaS from a sunny park bench. Apple Vision Pro + @Tailscale VPN + @cursor_ai handling vague prompts flawlessly over 5G. Gesture controls actually WORK. Remote dev setup of the future.

haha, that's crazy. I just built a Star Trek holodeck in @cursor_ai in 10 minutes of prompting with 03-mini. I have zero experience with threejs and I didn't have to create any models or assets, its pure code. I think this could be generated in real time using advanced voice

Holy shit. For the first time, I managed to build a complete feature, end to end, using the @cursor_ai Agent. I started with frontend components by copy pasting screenshots from Figma into the chat. Then I integrated the components with various pages, then wrote the backend and

How do you like the video? 😄 @FathomDotVideo @HeyGen_Official

I use Gemini Pro + Vertex AI SDK in TypeScript, but this works with OpenAI, Claude, or other platforms.

To scale: Add retries for failed attempts. Handle rate limits. Parallelize requests. Once this is stable, you can adapt it to any task with minimal effort! The crazy part: You don’t even need to rewrite the code yourself. Tools like Cursor AI can adapt the entire implementation

Example: For pricing data, I gave the LLM 3 tools: 1️⃣ Run a Google search via the Custom Search API. 2️⃣ Fetch content from a URL as markdown. 3️⃣ Submit the final answer. The agent iterates until it finds the info, retries intelligently, and submits precise metadata.

With function calling, you typically: * Set up a system prompt. * Add data/context in a user prompt. * Define functions. But this method often fails at scale—too much tweaking, unreliable results. Instead, use iterative prompting! Ask the LLM to take a small action (e.g., use a

I finally figured out how to delegate work to AI agents effectively. The key? Unlearning one-shot prompts with function calling and switching to an iterative loop. Now I can reliably offload tedious tasks, like finding pricing info for 30k. Here’s how 👇

Making this clip was an interesting learning experience in AI video generation. I tried Grok, GPT-4o and Runway, but it was really hard to get consistent images. Finally @Kling_ai gave me the best workflow: Generate images based on previous images, then "bring to life"