LLMs are cool

🇨🇦
Joined July 2016
Ode to @GrugBrainedDev grug sit by fire, code in hand, wonder why things not go as planned. complexity demon lurk in code, make grug heavy, big brain load. test shaman chant, "test all night!" grug just want code work right. shiny rocks spend on agile talk, grug prefer just take a walk. microservice far too much, grug prefer monolith touch. grug write log, many line, hope debug in less time. grug love tool, debugger dear, help make bug disappear. grug code simple, clear and bright, fight complexity demon with all might. grug say to young grug crew, "keep it simple, this much true. too much thinking, brain will crack, write good code, no looking back."
1
5
1
48
4,150
AI is a great tool.....
1
1
94
Monarch Wadia retweeted
My dear front-end developers (and anyone who’s interested in the future of interfaces): I have crawled through depths of hell to bring you, for the foreseeable years, one of the more important foundational pieces of UI engineering (if not in implementation then certainly at least in concept): Fast, accurate and comprehensive userland text measurement algorithm in pure TypeScript, usable for laying out entire web pages without CSS, bypassing DOM measurements and reflow
1,315
8,012
2,661
64,487
24,588,724
Excel is a cognitive crutch, I should use a calculator. No, wait, a calculator is a cognitive crutch, I should use an abacus. No, wait, an abacus is a cognitive crutch, I should use pen and paper. No, wait...
This Post is from a suspended account. Learn more
1
73
Monarch Wadia retweeted
The Turing test for AI video should be fixing the last season of Game of Thrones If we can do that, we can do anything
252
178
29
6,815
440,969
This is wild. I created a @trpcio-style router pattern, but for @electronjs IPC calls between the webview and Node.js. It took less than an hour using @claudeai . The type system is fully inferred from the Node.js code, just like in the original TRPC project. WTF.
84
Monarch Wadia retweeted
how my codebase written entirely with claude code runs
684
3,045
366
62,081
4,395,915
A new type of illiteracy: The afflicted are people who believe they can read well because they once did. They write well, and their brains r making fast connections and "saving time" so they skim really fast, lock on to key words, make connections about what they assume is being said due to pre existing assumptions about the author. then emotional and eloquently react to an argument you did not make. This type of illiteracy feels very insidious and scary because it's effecting academics and engineers and people society relies on
426
443
179
6,562
527,627
the 'boredom' experienced while waiting for the LLM to finish is often accompanied by anxiety, caused by not knowing exactly what the LLM-generated code does.
69
Coding without LLMs just feels primitive now. Exactly the same feeling I got when I started using Intellisense more aggressively. Can't go back now.
39
I'm blown away. Nano Banana 🍌Pro. I generated the 4-panel script using GPT 5.1, then fed it into Gemini. This was literally the first image generated. One shot, one kill.
338
Would not be using Rust on a serious project I'm working on, if I didn't have an LLM to help me get over the learning curve.
38
Gen Z-ers who are "raw-dogging boredom" to fix their attention spans is nothing new.... people have been doing it for millennia. It's called meditation.
56
To get to flow states while the LLM is working, think about the product design and the architecture. Not the code itself.
105
LLMs use Markdown for formatting. It's by far the easiest way to write formatted documents using AI. Maybe we'll see the full Markdownification of the corporate world? Maybe even some LaTeXification. The thought fills my heart with joy.
1
232
In the morning, I wrote Rust. In the afternoon, I wrote Python. In the evening, I wrote Typescript. At night, I had a prototype.
1
68
Immensely enjoying Rust so far.
93
we thought we'd find little green men. instead, we found a little green fenn.
64
It's too easy to hope that your LLM will fix your code by itself. It won't. If you start drifting away from deep understanding of what your code does, you're starting to lose power.
63
There's a new "modality" to LLM-powered development that i'm discovering. i can "brute force" code by endlessly adding unit tests that describe the expected behaviour in increasing detail. the code is lower quality, but delivered much faster and is "good enough."
1
116
One more advantage of this modality: you can ask LLMs to write specs. For example, "read my existing unit tests and tell me which edge cases are not covered. then, write tests for them." This makes it very easy to reason about & implement code.
1
58
Sample output from the LLM while working in this modality.
41