Ian
ianamo.bsky.social
Ian
@ianamo.bsky.social
Dev working with Python and React. Interested in edtech, AI, classics, language. An expert on explaining things and measuring how badly it went.
You know, Apple really undersells the Shortcuts app. It’s basically a whole visual programming language for hacking your iPhone.
January 10, 2025 at 1:13 PM
The ASI benchmark we actually need is when an LLM can convincingly recreate Aristotle’s lost treatise on comedy
January 7, 2025 at 12:27 AM
Next.js and prisma 🧑‍🍳👌💋
December 26, 2024 at 3:24 PM
Reposted by Ian
Ynterviewer: Kan you explayne thys gap yn your resume?

Dante Alighieri: Yn the middle of thys journeye of lyfe Ich founde myselfe yn a shadowye woode, for...
December 21, 2024 at 7:31 PM
Best way to learn is always to wade in and start hacking. ...With some prudent assistance from ChatGPT, of course.
December 19, 2024 at 2:03 AM
We would be rescued from an astonishing amount of nonsense if people only remembered that probabilities must be multiplied, not added.
December 15, 2024 at 9:54 PM
Saying you don’t like the taste of tofu is like saying you don’t like the color of blank canvas.
December 15, 2024 at 8:21 PM
Using LLMs to generate useful regexes > all other uses of LLMs
December 13, 2024 at 11:31 AM
Small is the new big
They have found that LLMs' capacity density grows EXPONENTIALLY over time!

Based on evaluation, they are seeing capability density DOUBLES every 3.3 months, which implies that in 3.3 months, we can achieve performance comparable to the sota LLM using a model with only HALF the number of parameters!
December 8, 2024 at 3:22 AM
Human beings: haha this AI program someone made last week can't do hands right
Also human beings: *can't draw shadows correctly after centuries of practice*
thereader.mitpress.mit.edu/the-art-of-t...
The Art of the Shadow: How Painters Have Gotten It Wrong for Centuries
The goal is not to expose the “slipups” of the masters but to understand the human brain.
thereader.mitpress.mit.edu
December 7, 2024 at 3:17 PM
Let’s tap Jeff Foxworthy for a new game show: “Are you smarter than an LLM?”
December 6, 2024 at 11:44 AM
React can be useful, except it requires you to do routine things in fussy, idiomatic ways
December 2, 2024 at 9:41 PM
Reposted by Ian
Did you know that attention across the whole input span was inspired by the time-negating alien language in Arrival? Crazy anecdote from the latest Hard Fork podcast (by @kevinroose.com and @caseynewton.bsky.social). HT nwbrownboi on Threads for the lead.
December 1, 2024 at 2:50 PM
Little known fact: the guy who decided that an http request between two ports on the same server should trigger CORS policy also designed Hawaii’s interstate highway system. Same guy.
November 30, 2024 at 7:18 PM
Parts of this are difficult to listen to, but there are also some very interesting reflections

youtube.com/watch?v=JPpk...
YouTube
Share your videos with friends, family, and the world
youtube.com
November 29, 2024 at 11:46 PM
Only noob cooks just follow a recipe slavishly. They get inspiration from what others have tried, and add their own twist to it, adapting to what they have on hand. When it comes to programming, LLMs are best used in just the same way.
November 29, 2024 at 5:44 PM
Spending money on possessions is also “consumerism” but not spending money on travel. A curious category!
November 29, 2024 at 5:17 PM
It is interesting that buying gifts is coded as “consumerism” but buying fancy things to cook isn’t.
November 29, 2024 at 5:17 PM
Python: Here's some function annotations!
Me: Groovy! So these will change the behavior of my function somehow?
Python: Certainly not!
November 28, 2024 at 5:29 PM
Data 👏 is 👏 not 👏 a 👏 crime
FYI, here's the entire code to create a dataset of every single bsky message in real time:

```
from atproto import *
def f(m): print(m.header, parse_subscribe_repos_message())
FirehoseSubscribeReposClient().start(f)
```
November 28, 2024 at 3:24 PM
“You can look at me walking in public. You can even record me walking in public. But, so help me, if you record me walking in public and then use that data to teach a robot how to walk, I will sue you out of existence.”

???
November 28, 2024 at 12:45 PM
“I am going out in public but robots aren’t allowed to look at me” doesn’t really sound viable long term.
November 27, 2024 at 10:53 PM
Writing a backend API in Flask is so beautifully, beautifully simple.
November 27, 2024 at 10:10 PM
Have I gone too far this time with my LLM experimentation? You be the judge! ianamo.github.io/2024/11/27/e...
The Ethics of Using AI… To Write Your Thank-You Notes?
We’re well accustomed to an ongoing conversations about the ethics of AI — the ethics of using other people’s data to train AI, the ethics of using AI on academic assignments or even legal briefs — bu...
ianamo.github.io
November 27, 2024 at 9:59 PM