
"Trust Me Bro" Benchmarks — How AI Labs Grade Their Own Homework
Why I Wrote This Every few weeks, a new AI model launches. The announcement almost always...
Tag archive

Why I Wrote This Every few weeks, a new AI model launches. The announcement almost always...

Here's how a lot of prompt changes ship. You tweak the prompt, try it on three examples you remember,...

I ran a fresh eval last week and the headline number came back 88.64 — ahead of whatever sat on the...

Originally published on vibki.com. Have you ever changed a prompt, liked a few answers, and shipped...

Note This is a shortened version of the original writing on kig.re — my personal blog. Here is...
A field guide to the handful of scorers that catch real LLM regressions, from exact match to LLM-as-judge to voice-specific misheard-number checks.
Change a prompt to fix one case and you can silently break five others. Nothing throws, no test goes...

Ricursive Intelligence Ripple-65B-Preview: New Dense Model! Context Ricursive...

Like GPT-6 Astra, Qwen 3.8 27B can interpret model weights Context Like GPT-6 Astra, Qwen...

Can a model hide its ability? Sandbagging research, a paired-task example, and checks that separate strategic underperformance from ordinary failure.
!["[Paper] Sparks of Weight-Space Programming: LLMs can natively read & write neural network weights"](https://media2.dev.to/dynamic/image/width=1000,height=420,fit=cover,gravity=auto,format=auto/https%3A%2F%2Fraw.githubusercontent.com%2Fhabibzuhaer%2Fp2p-buffer-assets%2Fmain%2Fdevto-assets%2Fcover.png)
[Paper] Sparks of Weight-Space Programming: LLMs can natively read & write neural network...

LLMs are not superhuman at reading and writing neural network weights Context LLMs are...