My first Chrome extension just got published! 🎉 It’s a simple tab manager inspired by Superhuman's keyboard-first approach I built it for myself and use cases, but I will be adding more features soon, so give it a try if you can and share your feedback https://t.co/nNOtUNTYkR

Bad: Burns credits quickly: this task consumed ~500 credits, so only 8 of this type and only 3-4 if more complex with their basic plan. The images did not match the phone for my task. Here is what I built: https://t.co/siXnRuVHFS PS: this was their 1.6 lite plan and not pro

Bulit something on @ManusAI: Good: • One of the best UX experiences I’ve seen in an agent, I could see what was being worked on without getting lost • Gave me a working, useful build inthe first attempt with a relatively vague prompt • Not the usual AI generative design

2026 interview questions: - you are in the middle of a refactor and the model says 8% context left before auto-compaction. what do you do? - how do you decide which tasks to give to claude, codex gpt 5.2 xhigh, and chatgpt pro? - tell me about a time you disagreed with an LLM

I'm using @crawl4ai for one of my projects, and what a beast of a library it is. The documentation is good, too. The open-source repo is well-maintained as well. https://t.co/CoetBRKK7b 💯💯

I've been using @antigravity for a few days and I am loving its agent-planning and reviewing interface. It offers clear visibility into plan modifications and the ability to provide inline comments and iterate, making it super powerful.

Just finished watching this on my commute. If your career is connected to technology, this documentary will definitely spark a lot of emotions. Brilliant storytelling. https://t.co/7z6HJHqtjw

"Give AI a task, and you get a task completed. Give AI a worldview, and you get something that thinks." There is much to learn from the work and knowledge that @koylanai shares here. Thanks for sharing your wisdom.

contd.

A fun Nano banana experiment: A user manual in comic book style. Here is one that I generated for my unpublished Chrome extension "Tab Triage"

Found a way to make nano-banana-pro insanely cost-effective on @fal I built a workflow that generates 9 4K images for a total of $0.21. That’s just $0.023 per image. That is a ~92% cost reduction. The key is grid generation. I added a 'Crop Node' as a workflow utility

“The Group description is system prompt” Wow this is a crazy setup.

A fascinating report derived from token usage data from OpenRouter, with many great insights. Of course, I had to use NotebookLM to digest it. https://t.co/TDZ4tbjZ6J

Weekend project: built a custom colouring page generator for my niece Keeping the same character consistent in an image showing both monochrome and color versions was harder than I initially expected. Doing it on Nano🍌2.5-flash instead of 3-pro made it even more interesting.

This could be one of the most powerful features in @magicpatterns and probably across other solutions for building and iterating complex UI and interactions. 🚀🚀

I frequently ask Claude to generate skills, slash commands, and agents for my workflows. The ability to use the tool to expand its own capabilities is incredibly powerful.

��Audience framing shapes how the model communicates; persona architecture shapes what reasoning pathways the model activates during generation. “

Google is shipping like mad!!!

Generated my first AI video for my sister’s venture with Veo 3.1. Sure, it needs refining, but seeing a real-use video emerge from one image and a prompt generated by Gemini in minutes... that feeling is something else. Now, I understand why people describe it as addictive.

My current prompt workflow: record on @WisprFlow, copy the transcript, paste it into ChatGPT, ask it to "Refine the prompt," and then use it. It works, but I’m sure there’s a smarter way. How do you turn your raw thoughts into effective prompts? Do share your workflows.