rahulmax
https://rahulmax.com/notes/how-i-keep-the-ai-bill-down/
In their case, context management pays off more the tighter your window. Their gap between managing and not managing is 35.7 points of success rate at 32k and 2.7 points at 128k. My version of that was a rule I stick to, as much as I can. I checkpoint a session at about 25-30% of the window, write the state out to a PROGRESS.md and a JSON file of the requirements, and start fresh. This restart costs me 30 seconds, since a bloated session doesn't get any cheaper the longer you stay in it.
Also worth knowing that the models are Nemotron-3 and Mistral-Medium, not the frontier models most people here are paying for.
I used Behringer Xtouch Mini mapped to Lightroom. This allows tactile controls, and also lets you keep your eyes on the photo being edited, instead of the constant back and forth between the controls and the photo.
Typestax is where you can pick and choose different typefaces and adjust it, preview them in a website and blog format, and also export your tokens in different formats.
I read this circa 2010 and it stuck with me ever since. Also, have felt this way over the years, when I see Windows users work.
- The eyebrow one is a straight bug and I'm glad you caught it.
- Base range 8 to 32 is excessive, agree.
- Per-element overrides existing doesn't help, if the thing that loads first establishes uniformity. I'll fix this too
- About the the tracking, you're right that flat -0.02em across h1 to h6 is wrong. I think the proper fix is to make it size-aware
- I'll fix and curate the pairs more, and aim for half of what I have now.
I was still in college at the time and used to be fascinated by how painstaking it was to build such UI in photoshop, and then putting everything in nested HTML tables to achieve this. Everything was rudimentary, CSS was in its nascency and everything broke in Internet Explorer 6
I'm a product designer and a design engineer and I love tactile interfaces and typography.
My goal was to challenge myself to build this purely by 'describing' the interface to an LLM, without giving it any image references. Being mosly working in Figma / Sketch for the past decade or so, the process was utterly painful and I craved to go back to Figma to polish and correct the interface, but I persevered and described the UI to the LLM and progressively built the UI.
The stack is Next.js / React / Zustand with zundo for undo/redo, Clerk for auth. I also made the LLM write 60-odd unit tests and playwright tests running against production. It was AI-assisted throughout.
chrismorgan's critique below is mostly correct. Thanks a ton! The eyebrow one is a straight bug. Base range 8 to 32 is indefensible. Word spacing shouldn't be a control at all. Those three are getting fixed today.
Midway through, the article stops being about Blade runner and becomes "people building things with AI are creating garbage", and Blade runner title cards is now a prop that substantiates that opinion.
It's fine if the typographic analysis holds up. It does not. I'm not able to see what OP is seeing. What am I missing?
I did the exact + fuzzy pattern on a project and it was so satisfying to watch people finding it useful.
Try it here. https://trustedgyn.com/ (a directory, for public good, hopefully you never have to use.) Make mistakes in the search bar and the model fixes it for you. For folks who are not native english, I found it helps a lot.
Like you said, working and maintainable are very different things. One-shot hits a wall the moment you need to do anything non-trivial after the initial generation. Bug fixing is extremely hard, even with AI assistance. Same with feature additions. It's pretty much black box at this point on. AI that wrote it now goes in loops wasting tokens without being able to can't reliably fix it either, because it has no memory of the architectural decisions it made (or didn't make, for that matter) the first time round.
What I realized is that the failure here is the absence of a shared mental model between you and the code.
I'm a product designer with average front-end know-how, and a solid understanding on HTML/CSS and how the web works, coming from the era of hand-coding html/css files. After vibe-coding a few products early this year, purely to learn how AI works, how to design AI interaction patterns etc., I built something called Intent Model. (largely inspired by SDD / BDD.
Intent model is a structured, typed artifact (basically a JSON contract) that captures actors, entities, journeys, rules, and constraints before I write (or make the AI write) any code. It sits upstream of everything. Think of it like a condensed, strict distillation of your PRD / BRD / requiremnt doc.
When you hand the AI a well-defined intent file instead of a vague brief, this one-shot becomes structured and bound by rules. Now you're giving it an architecture and to conform to. You define (or make the AI define) the precise variable names, their types, lifecycle, user roles, responsibilities, business rules and constraints in the file. Every generated artifact can trace back to a decision you made deliberately, reviewd and signed-off.
In the design world, we already do this by using design tokens. We can tell the AI that it needs to strictly use design tokens and not use stray properties like a hex color value or raw values not defined in the token contract. This is easily auditable by AI as well.
The result is you can still move absurdly fast and still maintain the understanding, which the one-shot approach throws away. This way, you know why every piece exists because you defined the intent before the AI implementated it.
AI is the accelerant, and you're the architect. The intent is the blueprint you generate to guide/harness the AI.
The best part is, once you have an intent contract at the heart of your project, it becomes impossible to break things too, logically or experience-wise.
https://computersciencewiki.org/index.php/File:Abstract_hear...
Hi, I'm Rahul. I'm a User Experience and Interface Designer with a strong Front-end Engineering background and I'm based out of Bangalore, India.
Portfolio: https://www.behance.net/rahulmax
Resume: http://rahulmax.com/assets/resume_rahul.pdf
Blog: http://rahulmax.com
Email: rahulmax[at]gmail.com
There's another video from 1949 about 'How a Watch Works' — http://www.youtube.com/watch?v=508-rmdY4jQ