The supreme first principle

> it's all just data

It's #friday

Time to chill and build something

For today I want to build a humble calculator that does basic arithmetic with rust + gpui

This was a fun lil project

GitHub

"Competition is for losers" >>>>>

The big downside of competition is that people start shipping out of anxiety

"I need to build this quickly and ship before the competition"

The luxury of knowing you are unbeatable.... by refusing to compete >>>>>

You use your time to release something truly astonishing

The big difference is that they work because they feel they have to, you work because you feel like bringing the new thing into the world, if your work isn't perfect then you won't release it... no pressure

The reason that apple created such great products is because they weren't competing with anyone

They spend years on a product because they have their own standards to deliver

Something you will never hear in Apple

"We need to release X product quickly before the competition"

They are literally just cooking they dgaf about competition

My YouTube algo is peak

Ya we normally don't have the mental model that lets us comprehend exponential growth.

Rewatched the ending multiple times

Its already #friday

There shall be no projects for this Friday

I got caught up in something and now I'm too tired to code

Lets see what the hype was all about

The only real way to benchmark it is when you run into difficult problems you yourself can't solve

All the fancy websites don't really mean anything... they just trained the model on a better dataset

An epic open source contribution from meta

Quick update

I'm pivoting to two products

#friday is here

Today's project was an interesting experiment

When Copilot first launched, LLMs were relatively weak and slow. We were forced to train custom, small-parameter code completion models, and that became the industry standard.

But why are we still doing that? General purpose llms have become powerful and we now have inference providers that deliver 2000+ tok/sec

So, I decided to test a new approach.

I used gpt-oss-120b on Cerebras to build a tab-autocomplete extension for VSCode. Its as fast as copilot and sometimes faster (with reasoning enabled lol)

The major advantage here is flexibility. For instance, I was able to get the model to treat comments as prompts and actionable instructions, rather than wasting tokens (like copilot) auto-completing them.

The sky is the limit here. This approach could eventually replace chat interfaces, have a auto-complete.md, and make agentic decisions. Much more to explore in the future! 🤟

[Code]

Great ads worked because they were rare and hard to make

If everyone can make great ads then people will just get desensitized to it

Ads already have such a low conversion rate. The reason that huge companies like coca cola show ads is to subconsciously influence your decision making. They are maintaining a brand not trying to grow it.

So just because you can make an amazing high quality ad, it doesn't mean you'll get users.

I'm very skeptical on media generation. I'm willing to change my mind but only when I see it's utility in the future. Let the dopamine rush die out and let's see people use it then.