24 May 2024

29 links · Friday

Successful language model evals — Jason Wei

Everybody uses evaluation benchmarks (“evals”), but I think they deserve more attention than they are currently getting. Evals are incentives for the research community, and breakthroughs are often closely linked to a huge performance jump on some eval. In fac…

On Template-Based Feed Generation

A couple of days ago, I was demoralized by a minor Jekyll issue. I was thinking about generating a single Atom feed from two content types. Turns out the plugin provided by GitHub Pages can’t do this directly.

Solodevs and the trap of the game engine

Since I was 19 I’ve wanted to make my own video games, as in creating everything in the game myself, often referred to as being a solo developer, or solodev. But I was bad at committing to anything and I always thought that I lacked some specific skill such as…

Thinking Big and Small

Writing about the big beautiful mess that is making things for the world wide web.

New MacBook Setup

This is mostly a reference for myself. I had to migrate to a new MacBook amidst a very busy week and I realized I have already memorized most of these steps, but this time I took notes. It took me about an hour. Next time will be even faster.

:epic-handshake: Reorg half a seat to the left

Handoffs suck. They’re the worst part of building products. But many organizations seem designed to maximize the number of handoffs on a project.

Supply Chain Data Maturity

Spreadsheets Working with data in supply chain can be difficult. Many companies have system silos that fail to utilize cloud data warehouse services like Sno...

A Simple QR Based Food Ordering App

Why During my stay at a hostel in Bangalore, I noticed the inconvenience of the existing food ordering system in the hostel. To order food, we had to call the cook and place our orders verbally or find the cook and place the order directly if you are new to th…

What is a collision?

From Mario bouncing off a Goomba to two cars bumping into each other in a racing game, dealing with collisions is such an integral part of most video games that we often take it for granted.

Enhancing Enum Handling in Spargine: Beyond Enums and into Versatility

This content explains the usage of Enums in programming, cautioning against relying solely on Enum values for human-readable names from databases due to potential performance issues. It introduces …

Clean Architecture Sucks

A brief conversation about the Clean Architecture approach and why some teams struggle with it.

Attack Techniques: Full-Trust Script Downloads

While it’s common to think of cyberattacks as being conducted by teams of elite cybercriminals leveraging the freshest 0-day attacks against victims’ PCs, the reality is far more mundan…

Attack Techniques: Remote Control Software

In yesterday’s post, I outlined the two most successful (and stupid simple) attack techniques that you might not expect to work (and you’d be so very wrong): “Please give me your …

We’re Ending Our Samsung Collaboration | iFixit News

It’s not us, it’s you. It’s with a heavy wrench that we have decided to end our partnership with Samsung. Despite a huge amount of effort, Samsung’s approach to repairability does not align with our mission.