Allie K. Miller

The annoying part about being able to build ANY interface at the drop of a hat with Claude Code is that… You want to pull your hair out when a website’s design doesn’t serve you. Pinterest: why aren’t ads pinnable? Why is it so hard to see what is public or private? Why does the minus operator not work? Why do all the results look like the same 10 outfits? In the same way I want to refresh a ChatGPT response, I want to get a fresh set of pins that is randomly sampled. Qantas: why don’t you let buyers filter by time? Every other site - including Expedia - allows you to pick actual takeoff and landing time windows. Yours is ‘afternoon’ or ‘evening’ filters. Im CK and flew Virgin instead because of Q’s search experience. 💡 Do we think websites will have dynamic designs? Will they be controllable by the user? Or will websites go away because of agents/tools?

What makes for a multi-agent super user? I chatted with someone at the AI labs who described the most successful agent users as having “multithreaded brains.” Are people with ADHD better at managing multiple agents at once? Watching all the power and capability shifts (like who became a good prompter), and it feels like another is coming.

I am so very obsessed with the new Claude layout.

I’m gonna be honest about Clawdbot. I’ve seen dozens of tweets on it, watched tons of videos, read whatever docs and guides I could find, and I’m with @omooretweets on this one. I’m glad tinkerers are loving it. I’m glad people cleaned up 10,000 emails or coded entire apps in their sleep or ElevenLabs’d their way into a dinner reservation. The killer feature for me is actually none of those things. It’s much simpler. It’s text. My phone is my primary device. Computer is secondary. My most inspired moments are when I’m on a walk or gazing out a car window, not when I’m slumped over some hotel desk chair. I’m constantly on the go, in an uber, at the airport, running to a meeting, grabbing 5min free here and there. I am dictating non-stop through all those moments. Always dictating. And so… I want to phone-in to Claude Code. Alas, as a decently technical non-engineer, I just don’t want to spend hours/days setting something new up right now (I can also appreciate that many in my same position have set it up and love it). I’m in a big dive deep moment with Claude Code and multi-agents, and I’m happy. Also, I still like manually reviewing and approving high risk tasks (like Claude grabbing a random GH repo). Maybe I’ll change my mind. Maybe one of you will change my mind. Or maybe Claude/ChatGPT releases all of this soon and I won’t have to wait. But if you’re one of the 2M people that follow me for my AI business takes, it’s this: I think Codex is built for SWEs atm, I think Clawdbot is too technical of a lift (and agree with @gregisenberg that someone could make bank bringing this into SMBs or startups), I think Claude Code is perfect for gung-ho lightly to heavily technical business users, and I think Claude Cowork is a great start for less gung-go but still very AI-curious non-technical business users. And I think all my answers change in 2-4 months anyways

Gemini 3, are you freaking kidding me. Vibe coded a sign language recognition app with video enabled, confidence scores, sampling settings, and tips in under 5 minutes. Code written with Gemini-3. Gemini-2.5 handling detection. Video is sped up 2x.

The one feature in Claude that would give me a step function in my work would be scheduling. Could be Claude Code or even just a Claude Opus 4.5 prompt with skills. Massively prefer in web app over terminal. Similar to scheduled tasks with ChatGPT Agent Mode. Anyone have an easy way to do this? Cc @AnthropicAI

More non-engineers should be learning to use Claude Code. Two things stopping them: 1) it’s in a terminal and terminals are icky for business folks (if you say “omg no it’s so easy”, it’s probably because you’re a dev) 2) Most use cases I’ve seen shared online are either too complex or can be accomplished just as effectively with web-based Claude ai TLDR it feels inaccessible. BUT - for the brave and early adopters among thee - there are specific use cases where Claude Code offers distinct advantages, particularly for working directly with files and systems *on your computer* without uploading anything to web applications. Examples from people using it: - analyzing and batch-translating audio files - improving image quality across hundreds of screenshots - the growth marketing team at Anthropic built workflows that process hundreds of ads, identify underperformers, and generate new variations within strict character limits using specialized sub-agents Most of those still skew technical. Here are crazy easy examples for non-engineers looking to start: - create a new document in x folder - edit said document and save a new version - create a fake csv of data, ask for 3 HTML files with 3 different presentations of it (interactive dashboard, slides, slides in McKinsey style) - ask why the heck your computer is slowing down or acting weird (ex: checking Activity Monitor, memory, disk space, and suggesting specific fixes) - review your document and file management structure and get recommendations for improvements - process computer files with batch updates like rename/reorganize/reformat hundreds of files following specific rules (add “_FINALFINALv3” to the end of each file name in my client presentation folder) - extract insights from a folder of local files without uploading them (ex: review all contracts in my contracts folder and pull out key risks or renewal dates) - analyze content across multiple files (ex: review all presentations in my client pitch folder and identify gaps in sales messaging) The core advantage: you’re working with systems and files locally. What I have been saying for over 5 years and just now really starting to feel because of Claude Code: AI is an OS. These systems will undoubtedly become more accessible over time. For now, Claude Code is still mega calibrated for engineers or the very early adopters within non-technical groups. The vast majority of engineers in silicon valley have been using these systems for months. I’m advocating for non-engineers to start experimenting NOW. If you guys want a crazy easy to follow tutorial on setting up Claude Code on a Mac, similar to what I created for Claude computer use, let me know in the comments. (And I’d maybe also include my top tips so you don’t have to memorize all the command options )